Good afternoon, everybody. Welcome to today's
MIT Starr Forum: Power and Progress:
Our 1,000 year Struggle Over Technology and Prosperity. I'm Evan Lieberman, Professor of
Political Science and Director of the Center for
International Studies, which is hosting this event,
and I'd like to thank all of you for joining us today. So the title of the talk
is taken from the latest co-authored book by my colleague
Daron Acemoglu, an Institute Professor here at MIT. And the questions raised
in this terrific book could not be more timely. What is the value
of new technology for the human condition? Should we faithfully trust
that labor-saving ideas will make us all better off? How might social and
political institutions mediate the effects
of these new ideas? As in his other terrific
books, in this one Acemoglu once again
reveals a great sensitivity to the importance of
social and political power in structuring how we live
and exist on this planet, and in a context
in which generative AI and other technologies
are diffusing with such incredible
speed, we really need to listen to
what he has to say. So before we get
started, I'd like to remind everyone that the
book, Power and Progress is for sale in the lobby. And I know that a lot of
you have already done that. The line was great, so
the sales have been good. Please pick up a copy,
and after the event please feel free to bring it up to
the stage for a book signing. In addition, as per
our custom, we're going to conclude
with a question and answer from the
audience, and we'll please ask you to line
up behind the microphones to ask your one question. And I have an emphasis on
one and on question-- which is not always heeded, but
I will try as moderator to enforce that. So without further ado, let me
introduce both today's speaker and our discussant. So Daron Acemoglu really
doesn't need introduction, but I'm here to do
that, so I will. He's the Institute
Professor at MIT, an elected fellow of the
National Academy of Sciences, the American
Philosophical Society, the British Academy of
Sciences, and a member of the Group of 30. He's the author of six books,
including the New York Times bestseller, Why Nations
Fail: Power Prosperity and Poverty, which was
joint with James Robinson. The Narrow Corridor:
States, Societies, and the Fate of Liberty,
with also James Robinson. And this book, which he's
going to talk about today. His scholarly research covers
a very wide range of areas, including political economy,
economic development, economic growth, technological
change, inequality, labor economics, and the
economics of networks. Our discussant today, who I am
sure did not rely on ChatGPT for her comments-- right, they
are completely original, OK, good-- is Fotini Christia, the MIT
Ford International Professor of the Social Sciences, and
my colleague in the Department of Political Science. She's also the faculty affiliate
of the Center for International Studies, where she's Director-- well, effectively, of the Center
for International Studies, and she's also the Director
of the Sociotechnical Systems Research Center
at IDSS, and she's Chair of the Doctoral Program in
Social and Engineering Systems at the Schwarzman
College of Computing. So I'll look forward, as
I'm sure all of you will, to her comments. But, first, please join me
in welcoming Daron Acemoglu to the podium. [APPLAUSE] DARON ACEMOGLU: Thank
you very much, Evan. And thanks, Michelle,
for organizing. This it's a true
pleasure to be here. And I would like to point out
that my partner in crime here, Simon Johnson, is
somewhere here. Oh, here he is. Yes. So we have the double bill here. So it's my pleasure to be here
to share some of the ideas from our new book. And here is the
book, and the title. And I think today, after
so much hype and discussion about advances in AI,
especially generative AI tools such as
ChatGPT, I don't think we need to give a big
introduction that there are tremendous and consequential
changes in technology. But part of the reason why
we have written this book is because we think there are
some critical questions that need to be asked whenever there
are new technologies, which is often in human
history, and those are about the control
of technology. Who controls technology, and
how that actually shapes who will benefit. In fact, for transformative
tools such as generative AI, these are particularly
important because there are so many different directions in
which these technologies can be developed, and
it is quite possible that they could bring
broad based benefits or that they might
actually enrich and empower a very narrow elite. In fact, throughout
history, there are examples of very
consequential decisions being made following the
visions of powerful agents. Today, those powerful agents
are the optimistic technology leaders in places such
as Silicon Valley. In the past, they may have
been different people. For example, this gentleman,
Ferdinand de Lesseps, was both a leading technologist
of his day, and perhaps the techno-optimist of his
day, because of his big belief that the world had
to be opened up with big investments in
public infrastructure. And when other people thought
that this couldn't be done, or couldn't be done in the way
that would really enable ships to freely flow from
the Suez Canal, he single handedly
pursued that dream. He convinced others
to come on board, he called for big
technological advances when others thought that
his schemes wouldn't work. And he was very
successful, becoming one of the most famous
figures of the second half of the 19th century. But his belief in technology
and his own understanding of where technology
would go then made him blunder with
huge consequences in the Panama Canal,
where he completely ignored the science and the
conditions on the ground, and his schemes
completely collapsed, leading to the deaths of
more than 22,000 people. Now, these sorts of
larger-than-life characters having a huge impact on
where technology goes, again, is not something that we
are unfamiliar with today. But even those things
you might think are not so important
because you might imagine that whatever these
leaders of technology decide, there may be very powerful
forces that ultimately, almost automatically, are going
to bring broad based benefits. And, in fact, a very
critical part of that has to be through
the labor market. Most of us earn our living
through the labor market, and if any technology
is going to create any type of broad
based benefits, it must somehow lift
people in the labor market. Here, the economics
profession is very optimistic, in some sense. It is so optimistic, in
fact, that a very important proposition in
economics doesn't even have a name because it is
so much part of our canon. So Simon and I had to
invent a name for it, and we called it the
"productivity bandwagon," and it goes something like this. Technology improves,
meaning our capabilities for doing things, for example,
in the production process, gets better. As a result, productivity--
meaning how much we can produce-- increases. And this creates a series
of economic forces-- in particular, firms wanting
to go out and hire more labor because they have
become more productive-- and via that channel and through
the workings of the labor market, wages increase, and,
as a result, workers benefit. That type of story is the
way that most economists think about why it is that over
the last century, for example, as productivity went up,
wages and employment went hand in hand, and brought pretty
widely shared benefits. But if you go back
in history, you will see there are many other
examples in which things work out in a much
more complicated way. Here, we provide
pictures of two of those, one from the medieval
era, the other one from the early
industrial period. The one on the left is
a medieval technological breakthrough--
really revolutionary in terms of its conception and
what it did to the production process-- windmills that
massively improved capabilities in many sectors. But when you look
at the data, you don't see the windmills creating
this sort of productivity bandwagon that lifted
the living standards of the workers or the peasants. At the end, instead,
what you see is that a very narrow group
of people-- the landowners and the clergy who controlled
land and the production process-- were the
beneficiaries while the working conditions of farm workers
did not improve much. On the right, you have even
a starker technological transformation that, again,
has a more complex effect. Eli Whitney's cotton gin,
which, at one fell swoop, turned the US South from a
complete economic backwater to the largest producer
and exporter of cotton and fueled the critical phase
of the Industrial Revolution and really created huge
fortunes in the US South. But the workers who actually
did the cotton production-- the enslaved Black people--
did not see any benefits. In fact, their coercion
increased, intensified. They were moved to the Deep
South, where working conditions were harsher and longer. And, again, no sign of a
simple productivity bandwagon. But you might actually think
that those two are handpicked examples and the bigger
process that really defines our age-- the one
that started sometime in the middle of
the 18th century with the application
of scientific knowledge and artisanal knowledge
to the production process, the Industrial Revolution--
is very different. After all, when we talk
about the dangers of AI or other automation and digital
technologies for inequality for wages, one response
that Simon and I get is, well, are you saying that
this time is different? And, in fact, this is the reason
why the subtitle of our book is "Our 1,000 Year Struggle
over Technology and Prosperity." No, in fact, we're not saying
that this time is different. This time is very similar
to what went on in the past. There has always
been this tension about who controls technology
and whether actually the gains from technology are
going to be widely shared. So, for that, let's
turn to a re-evaluation of what happened in the
British Industrial Revolution. And the story turns out
to be much more complex than the simple story that the
Industrial Revolution increased our productivity and we are
all the beneficiaries of it. Yes, indeed, we are immeasurably
more prosperous, healthier, and more comfortable today
than people were 300 years ago, but, again, there was
nothing automatic about it. In fact, the path
to that improvement was much more circuitous. The early phases of the
British Industrial Revolution were characterized
by something that's going to have resonance for
today as well, automation. Meaning the application
of technology for simplifying and
reassigning tasks previously performed
by humans to machines. This was visible in
the textile industry, especially in the
weaving processes. And weaving, which used to
be done by people with hand looms or in their
houses, became something that migrated to factories. Did that improve productivity? Yes, the evidence is very
clear that productivity increased as a result of that. But, in fact, the
benefits of that, again, were not widely shared. The evidence is not completely
clear, but most of it suggests that real
earnings of workers in Britain during this
period did not increase, while their working
hours probably increased by about 20%. So their hourly
real wages declined. But even worse, the working
conditions of people were much harsher-- much
less autonomy, much less independence for
workers in factories than they had as
independent weavers. And working conditions are
only one part of the story, of course. People also care about where
they live, their health conditions, the
pollution around them. And in all of
these, the situation was decidedly not so good for
the British working people. So the factories were
emblematically much harsher places than they were used to. This is captured by-- symbolically, at
least-- by the picture we have on the left,
which is the idea of a very famous
economist Jeremy Bentham-- the panopticon,
famous from Michel Foucault's writings or the movies of
Guardians of the Galaxy-- but the way that Jeremy Bentham
thought of this was a highly efficiency improving technology
because it would enable employers to monitor
workers better, or teachers to monitor students,
or guards to monitor inmates better. And what could be
wrong with that? But, of course,
that was actually what employers were
very interested in doing in modern factories to the
expense of workers, who were forced to work very
long hours under very harsh conditions. At the same time, the cities
in which workers concentrated became complete cesspools,
much less healthy, much less comfortable living places. And life expectancy
during this phase at birth may have fallen as low as 30
years old, a terrible number at a time when
conditions economically were actually improving for
factory owners, for example. Now, this is, of course,
not the end of the story, and there is some truth to
people who say that, look, we have benefited so much from
the Industrial Revolution. Indeed, we have. But the process for which
needs to be understood, how it is that we benefited
from it after this early phase. And this early phase
was not a short one. It may have lasted
by about 100 years. The beginning of the Industrial
Revolution is not clear, but you may date it to around
1750, and, by the 1840s, conditions were
still very harsh, not just for workers
in textile factories, but in every sector of
the British economy, including coal mining, another
one of the dynamic sectors of the economy where
children as young as six were working 18
hours under hugely harsh conditions deep in mines. But, of course, you might
think, this is all history. Today is different. Yes and no. Today is different
because today we are in an economy in
which, for a while, we got used to a very different
type of sharing of the gains. And these two charts
that I have up here summarize both the ways in
which today is different, but also the ways
in which there might be some parallels to those
older, not so good times. What I'm plotting here is
for men and women separately and for five education groups--
starting from workers with less than high school in dark orange
all the way up to workers with a postgraduate
degree in dark blue-- how their real wages have
evolved over the last 60 years. And, in 1963, everything
is normalized to zero, so you can follow the cumulative
change of the wage profile for each one of these
10 demographic groups. What you see in the
1960s and early 70s is actually a continuation
of a trend that is also visible from other data
sources in the 1950s, which is one of shared prosperity. Real wages for all
10 demographic groups that I'm showing you here are
growing in tandem, more or less on top of each other,
from the fact that all 10 of these curves are
growing very sharply. In fact, they're growing
very, very rapidly about 2.5% every year in real term, which
is a really remarkable rate of growth. At that rate of growth, starting
from poverty, in two decades you can reach sort of a
much more comfortable level of living. But that period of rapid
wage growth and shared growth comes to an end sometime
around 1980 or the late 1970s. From then on, you see a
much different picture. These curves are
fanning out, indicating a greater level of inequality,
a much greater level of inequality. But even more jarringly,
you see that the real wages of low education groups,
especially men, but also for women, are
actually declining. So the green line, which is
for high school graduates, the orange line, which is
for high school dropouts, are actually sharply declining
from their values in 1980. Not only the gains
are not being shared, but some groups are losing out. So this picture therefore
poses two sets of questions. How was it that the US
economy-- in fact, it turns out, much of the
industrialized world-- reached some sort of
compact in which there was this fairly shared and rapid
growth in the late 40s, 50s, 60s and 70s, and why
did it come to an end? So let's try to
understand both of these. And let me skip this
part, which shows that inequality is also
increasing in other countries as well. Let me skip that,
and instead just get to the bottom of the theory
that Simon and I tried to develop in the book, which
is that the productivity bandwagon is not a force
of nature that applies under all circumstances
automatically and with great force, but
it is something that's conditional on the
nature of technology and on how production
is organized and how the gains are shared. In particular, what
changed in the second half of the 19th century in
Britain and then continued in the United States and much
of the industrialized world in the first part
of the 20th century, and came into even
greater fruition in the three decades that
followed World War II, was this process of
shared prosperity built on two big pillars. One is new tasks, and the
other one is worker power. Of course, both of these
terms need to be defined, and they may be a
little bit simplifying. But the new tasks are critical. Automation that I
mentioned, which is the substitution of
machinery and, today, algorithms for the labor of humans,
has always been with us, or at least has been with us in
great force since the beginning of the Industrial Revolution. Automation is a major force
for increasing productivity, but it does not create the type
of shared prosperity by itself. Because, after all,
what automation is about is to take tasks
away from workers and have more machinery do it. So it reduces the
importance of labor. It also reduces the
labor share in output or in national income. So if we're going to
have shared prosperity, automation needs to be
coupled with something else. And that something
else, critically, turns out to be new tasks. New activities in which
human labor is critical that reinstates workers centrally
into the production process. So throughout the second
half of the 19th century and the early 20th century,
we see these as very important determinants of what's going
on in the labor market. Emblematically, for example,
captured by the picture up there, which is from
Henry Ford's motor factories. Henry Ford was a leader in
applying new technologies, including decentralized
electricity and assembly line type technologies
early on, and that was absolutely revolutionary,
completely changing the car production, making
cars affordable for the masses. And a very important
part of that is the use of machinery to
do tasks that were previously done by labor, automation. But if you look at
the Ford factories, that wasn't the only thing
that they were doing. At the same time as they
were introducing automation technologies, they were also
creating new tasks for workers. So it is no surprise that,
in this picture, what you see is the advanced machinery
together with the workers. Workers are now performing
more technical tasks. They're operating
this machinery, they're engaged in design,
inspection, and other quality control activities. And if you actually
look at the factories of the early 20th
century, you'll see that production
workers are joined with non-production workers,
clerical workers, that are very much engaged in
planning and other aspects of the production process. It was this double
process that was so important for the beginning
of shared prosperity. Automation, which increases
productivity, but also new tasks that give another
boost to productivity and also create
reasons for workers to share in those gains. But even new tasks by
themselves are not enough, because if workers are
making a major contribution to productivity but they
don't have the power to take a share of
that, they may go down the same path as the
Black enslaved workers with the Eli Whitney cotton gin. It may not be in the
interests of the firms to share those gains
with them, and they can get away with it because
they have all the power. So, actually, a
balanced distribution of power in workplaces
and in society is also a critical part of it. That's why the second
picture on this slide is one of the emblematic
strikes in the auto industry that was still a leader
in establishing the labor movement during this period,
the United Auto Workers strike at General
Motors in 1937. So it was this double process
that was so important, but not just for the early car industry,
but, in the 19th century-- the second half of
the 19th century that I've been referring
to in Britain-- what was so
distinctive about it-- when you think about what went
on, the direction of technology changed. Completely new set of
technologies in railways, in steel, and chemicals
that were much more important for improving
the productivity of labor and introducing new
tasks for labor-- this was embedded
in an society that was democratizing from the
times of the early 19th century, where even the middle classes
didn't have the vote, now universal male suffrage,
and then universal suffrage came to Britain. And, also, the labor movement,
which was completely banned and heavily prosecuted
up to the last quarter of the 19th century, became a
staple of British workplaces and was a very important part
of improved working conditions and wages. Now, this was about why shared
prosperity's foundations were laid starting sometime in the
middle of the 19th century, and then continuing to the 20th. But then why did it come
apart sometime in the 1980s? And I think the same two
processes now playing out in reverse are the
key actors in this. And to explain that,
I'm showing here a modern car
factory, which looks somewhat different, or quite
different, from the Ford one. You again see the
advanced machinery. Now the advanced machinery
takes the form of robotic arms. But conspicuous in its
absence are the workers. The workers are no longer
playing a central role. The automation is
rapid, but the new tasks haven't accompanied it. So too much focus on
automation but not enough on creating new tasks is
the technological part of it. But accompanying that has also
been an institutional change. And that institutional change-- sorry, before that, let
me actually show you this figure to substantiate
the claim that I made. This is a figure
from work that I have done with Pascual Restrepo. What it shows is something
that's akin to the first graph that I depicted, the evolution
of the real wages of the 10 demographic groups. Now a little bit more
detailed, demographic groups distinguished by age, gender,
education, and ethnicity-- each one of these circles
refers to one of them. On the vertical
axis, I'm showing you the cumulative change
from 1980 to 2016. So that period in
which some groups were experiencing wage
growth, other groups were experiencing wage decline. You can see the same
thing from here, from the fact that many of these
circles are below the zero. Those are the
demographic groups that are experiencing wage declines. And on the horizontal axis,
I'm depicting the extent of task displacement that
a demographic group has experienced during this period. Namely, what
fraction of the tasks that this demographic
group used to perform across industries and
across occupations in 1980 have since been automated. So you can see that,
for some groups-- mostly those like us who
have postgraduate degrees or very high levels of
specialized skills-- those numbers are
very close to zero. We have not really suffered
much automation of the tasks that we used to perform
that are much more creative, much more problem
solving, and high level. But if you look at those for
high school education or less than high school education
demographic groups, shown, for example,
by purple and green, you'll see that up to 25%, 30%
of the tasks that they used to perform have
since been automated, and those are exactly the groups
that have suffered the wage declines. In fact, this regression line
explains about 60% to 70% of the variation in equality
between groups in the United States. So this is the automation part. But automation has had a
very big effect and, in fact, its path was very much shaped by
institutional changes in the US labor market. And those institutional
changes have been in the direction of
declining worker power. And to understand
declining worker power during this period-- in the same way
that if you wanted to get to the details
of understanding the increase in worker
power in the 19th century, you would need to think about
both how ideology has shifted and how organizations
have shifted. And both of those have
gone against workers during the post-1980 era. One is-- perhaps I'm
giving too much credit to our fellow economist
Milton Friedman-- but the rise of new
corporate visions which elevated managers doggedly
working for shareholders and ignoring everything else. This was the beginning of the
shareholders value revolution, or Milton Friedman's
statement that the only social responsibility
of business is to maximize its profits. And that was coupled with
the erosion of worker powers, for example, emblematically
during the defeat by Ronald Reagan of the
professional air traffic controllers strike. So these two changes
together both shaped the way that managers wanted to approach
how to run their business. For example, monitoring
workers more tightly or automating in order
to cut labor costs was completely welcomed because
it would increase the returns to shareholders. But also there was no resistance
to them from organized labor because organized
labor was getting weak during this period. Now, this is all before AI. Can the age of AI change it? Yes. The age of AI can change
it because, at some level, if you look at the details,
the promise of generative AI-- some of it is hype, but
some of it is reality-- is that it can actually be a
tool in the hands of workers. But if you look at
the reality of it, you also see major roadblocks
towards that kind of change. What makes today such
an important point in this type of
discussion is that there are transformative
and very consequential choices ahead of us. And, again, Simon and I think
that one way of framing this is in terms of
different visions. One vision about where AI and
digital technologies in general are going to go-- emblematically summarized
by that picture at the top, or the Turing test-- is
towards autonomous machine intelligence. Meaning machines become more
and more autonomous, more and more intelligent,
and they start doing more and more of the
tasks that humans used to do. It won't take much
imagination to see that if that's the
emphasis, we're going to have a lot more automation. But, in fact, if we are
right that automation doesn't create the foundations
of shared prosperity, that spells trouble. But that's not the only way in
which digital technologies were conceptualized, and it's
not the only direction in which AI can go. Long ago, many
computer scientists understood a very different way
of using technologies, which Simon and I call "machine
usefulness" to contrast with machine intelligence. The objective is
not to make machines intelligent in
and of themselves, but more and more
useful to humans. Engineers such as MIT'S
Norbert Wiener, JCR Licklider, who was briefly at MIT as
well, or Douglas Engelbart tried to articulate both the
philosophical foundations and the technological realities
of this vision, and out of this came many of the
technologies that we rely on. For example, when you use
your smart phone, the menus, or the computer mouse--
which was revealed by Douglas Engelbart in a
very famous event called the Mother of all Demos--
or hypertext, all of these came from an effort to make
machines more usable and more useful to humans. And, in fact, AI could
pursue that path. Now, the problem, in fact, it's
not just one of distribution. If you overemphasize
automation, it's not that you're going to
get huge productivity gains and they're just going to
be unequally distributed. In fact, there is every danger
that overemphasizing automation is not going to get you much
productivity benefit either. And this is the concept
that Pascual Restrepo and I and Simon and I tried to
capture with the label "so so automation." What you're trying
to do is you're trying to get
machines to do things that humans are pretty good at. So when you do that, you don't
get a huge productivity boost because humans
were doing it fine, but you get big distributional
benefit or distributional costs because you're
sidelining humans. Firms may become a little
bit more profitable, but a lot of workers lose out. And self-checkout kiosks or
excessively automated customer service, all of those are
examples of so-so automation where the productivity
benefits turn out to be not so much as
people were hoping. Now, of course,
generative AI and ChatGPT could change all of that. So we asked ChatGPT itself
whether generative AI could reverse these things. On this one, I think
ChatGPT was quite on target. Perhaps, but probably not. It's not a magic solution. If generative AI is used
to replace workers instead of support them, it could
have negative consequences. Now, we don't know whether it
knew the answer that we wanted to hear or it read
some of our papers, but we agree with this answer. Labor market consequences
and inequality are not the only things
we have to worry about. One of the other trends
since the 1980s but, again, accelerating
with generative AI, is about who
controls information. After all, even in the
Industrial Revolution, it wasn't just
automation, it was also how the modern factory system
changed the method of control and who was in charge and
what they could dictate. So one of the things
that we are seeing with more and more digital
technologies is surveillance. Surveillance in
workplaces, surveillance in political views. Now, it takes different forms. In China, you may be
more worried about it because it's in the hands of the
government-- the social credit system, or facial recognition
cameras everywhere where protests could
one day break out. In the United States,
it's companies. It's Google,
Facebook, Amazon that have all of this information. But, at the end of
the day, in the book Simon and I argue that both
of those are pernicious. It doesn't matter
who has your control. As long as that
information can be used without any
constraints, it's going to be
anti-democratic and it's going to be inequality inducing. This has so far been a lot
about the developed world. I have given examples from the
US, a little bit about Europe, and, in fact, that's a
national focus for Simon and I because we want
to trace the counters of new technologies
and how they are used. But let's not forget that
new technologies that are developed in the
United States and in China are going to be used throughout
the world and, in fact, the International
division of labor is already being reshaped
by automation technologies. One of the patterns
you see around the world is that a lot
of routine activities that were automated in
the United States are also being automated
around the world, or at least the amount
of activities or labor that is assigned to these
production functions is declining. There is every danger that AI,
if it goes down the automation path, could be a
highly an equalizing technology around the world. And, again, some of the
surveillance implications are global as well. The recent work by some
of my colleagues here-- Martin Barrera and
David Yang and others-- shows that Chinese
companies are already exporting anti-democratic
monitoring technologies to more than 60 non-democratic
countries around the world. So this is all
potentially depressing because it says there are
big dangers of [INAUDIBLE] inequality and democracy. But, from the beginning, I
tried to frame this as saying, well, these are
transformative choices because there is no
necessity that AI is going to go one way or another. There is a high
degree of malleability to all technologies, and
that's doubly true for AI. And if we make
the wrong choices, they could have
damaging consequences. If we make the
right choices, they could be much
better for society, it could be much
better for workers, it could be even
better for democracy. So the question
is what we can do. Simon and I, in
the book, suggest that we need a
three step process for thinking about change. The first one is
changing the narrative. Our modest hope is that this
book is a small contribution to changing the
narrative, moving away from blind technological
optimism-- everything is going to work out. This time was no
different-- to try to understand how things have
worked in the past, when things go right, when things go wrong. And part of changing
that narrative is to also recognize that things
are more likely to go right when there are more voices
rather than technology being in the hands of some
powerful actors, be it Ferdinand de Lesseps
or Sam Altman and Elon Musk. So the first is a
change in narrative. But changing the narrative
is not worth that much unless there are institutional
and other developments that actually turn that into actual
action and policy changes. So that is what we mean
by countervailing powers. So part of the reason
why things were different in the
1950s and the 1960s was because
technological choices-- and how those choices were-- how
those gains were being shared, was embedded in an
institutional framework where government
regulation was important and where there was civil
society and labor movement constraints on what
companies could do. Some of that needs
to be recreated. It has to be in the form of
a new labor movement, perhaps other forms of bottom up
organization for civil society, and also government regulation,
especially in the field of AI. And the central idea here
that Simon and I emphasize is that redirecting
technological change has to be a major part of both the
efforts of the labor movement and of government regulation. Technology has many potential
directions in which it can go, and there is no guarantee
that the completely unfettered market process is going
to choose the socially beneficial direction. There were many people, such
as this gentleman Ted Nelson-- not just Douglas
Engelbart and others-- who thought that the
personal computer and other digital technologies
would be fully liberating choices, both for
workers and for citizens. In the end, that's not the
path that we ended up on. But that was a choice. It wasn't because they were
completely wrong in thinking that technology could be
a decentralizing force, it could be a tool in
the hands of the workers, not just of corporations. And, in fact, Ted Nelson
very much anticipated this when he was writing. On the one hand,
he was optimistic, but on the other hand,
he was very much emphatic that large corporations
such as IBM would try to control
the technology, and that would push it in
a very different direction. But all of this raises
another question, which-- and I will end on this. There is some degree of optimism
in saying that we can actually redirect technological change,
and that's a social choice. Because the counterargument
is, no, technology is a fully organic process. Every time you
interfere with it, it's going to end
up in your face. Well, of course it is
an organic process. But the fact that it's an
organic process doesn't mean that it cannot be
steered within bounds. And one example
where you see that is in the energy
sector, which is not, of course, something
to be proud of. We are very much behind in
combating climate change. But, today, if you
look at our ability to generate clean
energy, it's miles apart from where it was in
the mid 2000s or even in 2010. For example, various different
types of solar and wind technologies are, today, cost
competitive with fossil fuels, whereas they were about 10 times
as expensive as fossil fuels as recently as 2010. How did that happen? It happened because
there were some subsidies and some regulations
about clean energy, and it was also a
civil society movement. More pressure from consumers who
wanted cleaner products, more pressure from civil
society for companies to clean up their act. And even a modicum of
that type of pressure led to a complete
redirection of technology. And the reason why we end
the book with this example is precisely because
Simon and I think this can be done in the realm
of production technologies with even more
consequential consequences than energy perhaps. Thank you. [APPLAUSE] FOTINI CHRISTIA: Hi, everyone. First of all, I want to
thank CIS for the opportunity to engage with Daron on this
terrific book co-authored with Simon. It's great to have
you in the room. It's a true honor to be here. As you saw from this
terrific presentation, this is a book about the
forces of technological change, but also all the challenges
that this ushers in. And, I mean, it's an
intellectual tour de force taking us over 1,000
years and bringing it to the very present,
focusing on the challenges of digital innovation
and the new era of AI. So, unsurprisingly-- and now
you know it's not just ChatGPT-- the first thought
that came to mind was how relevant these topics
were even in ancient times. So it's not just
the 1,000 years, but these were real
preoccupations also for the ancient Greeks. And in their myths and in
their philosophical writings, they were really
challenged by this idea of the wonders of innovation
and technological change and what they meant for
creating societal hierarchies and structures, but also
a lot of cautionary tales of the responsibility that
comes with these great wonders and these great capabilities. So, if you may
indulge me, I'll touch on three myths
that came to mind, and I hope I'll
briefly travel us back to the Greek
mythology books that we all engaged
in childhood. So the first is the
myth of Daedalus. He was a master
craftsman and inventor that was brought to
the island of Crete by King Minos in
an effort to house the Minotaur, the
half-bull half-man that was a pretty dangerous being. And, of course, the
innovation there was the labyrinth that enabled
the housing of the Minotaur, but it was also-- it created a big
inequality, and that was in the form of
the human sacrifices of the Athenian youth for
whom this became a death trap, the actual labyrinth. The second myth is
the related story of Icarus, the son of Daedalus. After this great craftsman
fell out of favor of the King, he was closed up, locked
up in a tower with his son. And, of course, he wanted to
engineer their great escape, and he created wings
out of feathers and wax. And while they were flying
out and were actually making their trip back, Icarus was
not listening to his father and decided to fly very close
to the sun, which actually led to his kind of plummeting
in the Agean Sea and dying. So what was interesting-- I mean, this is seen often
as a myth about hubris, and very much hubris around
invention and innovation. And, I think, thinking to
contemporary times and kind of youth and hubris, I thought
of cases like Theranos and FTX that I think are
quite associated with some of these startups. And the third and
last Greek myth may be the closest to what
we're discussing here, is Prometheus and his decision
to steal fire from the Gods and share it with the humankind. And, in that sense,
Prometheus was kind of the ultimate
equalizer, maybe a union organizer of
his time of the Titans, deciding to kind of
challenge elite dominated decisions over technology. And he took this great
leap, but he was also punished, very, very much so. I mean, he was chained-- eternal punishment,
being chained on a rock and having an eagle eat
your liver every day. And this would go
back, and every day he would get rejuvenated
and eaten again. So I wonder if there is an
interesting lesson there, kind of the courage
and self-sacrifice that may be required in
order to rein in some of these really big interests. And apart from myths, I
mean, Greek philosophers were also very preoccupied
with these themes. Plato, in Republic, talks about
a controlled technological environment where
the main goal is the pursuit of the
good life, implying that some of these
technologies are actually distracting from this goal. And then Aristotle, in turn,
acknowledging, obviously, the great importance
in society of having-- I mean craftsmanship
and technology-- but distinguishing
between technologies that sustain life versus
those that actually enhance living, which I think
is an interesting distinction. I mean, so this is
clearly a relationship that has been kind of a
great topic of importance throughout life,
and I was hoping to pose three sets of questions. I hope Evan will not block me. But this is kind of a bit of
a warm up for our discussion. I'll pose all three, and then
you can take them one by one. First, the book tries
to suggest that there is a certain heightened
energy and urgency about digital innovation
and AI that maybe did not exist with other past cases. But I wonder if
this may have been what it felt like at the
time for all the others, what it felt like at that
particular point in time. What is different, actually,
between the Gilded Age-- Rockefellers, Vanderbilts--
and the Bill Gates, Jeff Bezos, Elon
Musk of our times? Why is the invention
of railroads and steel at that time so different from
digital innovation and AI now? It seems like in every
past case in history-- and I don't know
if that makes me a great optimist-- but humankind
kind of, in the long run, benefited from this
technological change, managed to improve overall
standard of living, averted disaster,
moved on, and then went into a new cycle of innovation. So why and how is our
case now different? And I particularly
wonder about this in the context of the
work around climate crisis and the
environment, which you highlight very
articulately in the book, and which seems to give us
a certain kind of roadmap and recipe of how to think about
this and how to move forward. My related second
question is whether it is technology,
really, that creates the inequality, or lack
of proper governance over technology. And, specifically, this is a
question about understanding how technology
relates to societal, economic, political
structures, and basically kind of the role of the broader
ecosystem of the cities and the state and other
institutions in addressing these inequalities. So I wonder, for example,
what does a union-- an effective union-- look
like in the 21st century? And I'm particularly
interested in-- especially now, I think President Biden
just joined the picket line in Michigan where the
United Auto Workers Union is on strike against the
three big auto producers here. So what does it mean to
have an effective union in the 21st century,
and what does it mean for collective
action among citizens to be effective in this case? So there is a new book by
Frederick de Boer that I was just reading about on how
elites ate the social justice movement-- this is the
title of the book-- and he talks about the
failures of some of the most recent movements, like the
MeToo movement, the Occupy Wall Street movement, the Black
Lives Matter movement. He basically claims that
they may have succeeded in symbolic change among a
lot of the academic elites and some of the more
bourgeois citizens here, but they didn't really
manage to make changes for the average citizen. And he says that a lot of
it has to do with the fact that they didn't get any
legislation actually passed on these issues, and
he also attributes some of the failure in the way
these movements were actually structured, which is
very different than some of the movements that we
know from the 20th century, like the Civil Rights
movement, for instance. Beyond citizen
movements and unions, I was wondering about the
responsibility of state more actively in
terms of regulation and the creation of
social safety nets. I think there are some
people out there who may even argue that the state has
already been co-opted and what is
regulating is in favor of the interests of these
big elites and corporations and not really
actually regulating in ways that are protecting
people and citizens in terms of these inequalities. Also, what is the responsibility
of international institutions? We know Europeans are
very keen on regulating on these issues on the EU level. So can we think about this
differently in that context? Or academic
institutions, like here. I mean, you made it clear that
a lot of us that actually get money from these
big tech companies need to be very clear about
where our independence may potentially end
or be compromised, and I wonder, what is the
role of other initiatives, like ethnic and social
responsibilities of computing initiatives, or other
efforts within academia, to try to think about how to
move beyond these inequalities. And as the last question-- this is a little bit more
about the democratizing of technologies and this
idea that they have also flattened inequalities
to a certain degree. So, for instance, there are some
instances in the Global South where they have been faster
adopters of technologies than we have, and how the
openness of digital education, information has actually enabled
them to make big strides. They've also been very keen
on some of the AI health care tools, for instance,
that they've taken on in some of their
national health systems. And even among educators,
among creators, among artists one can see how technology
has been an enabler. And can we say that
maybe we can accept that there will be a certain
level of inequality that is required to keep wanting
to have change and move on towards progress,
and how should we look at this in terms of the
Global South and the Global North? DARON ACEMOGLU: How
many hours do we have? FOTINI CHRISTIA: Sorry,
I mean, literally just to open up some thoughts
of conversation. And we could even just turn
to the audience right away. DARON ACEMOGLU:
No, no these are-- Fotini, thank you for these
very erudite and extremely far ranging comments and
fantastic questions. But I will have to be brief. Let me first say
that Simon and I used to have Prometheus as the
beginning part of the book and try to tell the
story of responsibility. Actually Prometheus
himself is sort of a conflicted character,
hubristic and responsible at the same time. But then we decided A, it
was going to be a little bit tortured, and B,
we would probably be caught by somebody who knows
Greek mythology much better than we do. FOTINI CHRISTIA: And
then Oppenheimer, I feel, had already taken Prometheus. So there you go. DARON ACEMOGLU: Well, we also
had Oppenheimer and Szilard. We had the contrast
between Leo Szilard-- if we had known that the
movie was coming out, we would have kept that. FOTINI CHRISTIA: Prometheus. DARON ACEMOGLU: But absolutely. You're right. And that's part
of the reason, is we think these issues
are as old as humanity, that there were
concerns as early-- and even probably earlier than-- ancient Greek civilizations
about control of technology. So they are very much with us. Coming to your questions. It is possible to read
history by saying, look, we've had some hard times,
but then we bounced back. But Simon and I don't think
that's the right story, partly because A, there are many
examples in which we did not bounce back. Or even when we even
when we bounced back, it was far from automatic. I think you cannot tell the
story of the medieval Europe as one of bounce back. There was no sort of
process towards sharing of those gains with farmers
or with the farm laborers until the whole
system collapsed. And when you look at the
two examples that you hinted at-- the British Industrial
Revolution and the rapid industrialization with new
industries in the United States-- in both cases, there was a
very, very radical change in institutions, and
that was very far from a foregone conclusion. The progressive
movement, I think, even its participants
were surprised that it could actually form
that coalition and succeed. And the cards were
very much stacked against democratic reform,
the labor movement, and all of that in Britain. One example we mentioned
briefly in the book because it's so telling,
that, in the 1840s, the Chartists in the UK
collected 3 million signatures. I mean, can you believe
3 million signatures in a time when you don't have
any of the modern communication technologies? With the demands that
are so tame in terms of universal suffrage and
some basic sort of rights. And they were so careful not
to be labeled Socialists. And the response of
Parliament was completely turn down all of these demands
and jail all of the Chartists. So there was no automatic
institutional process that could say, oh, yes, we're
going to bounce back from this. So that is the
sense in which there is both caution and optimism in
what Simon and I are telling. The caution is, there's
nothing automatic here, but optimism is, yes,
it is possible to do it, but it's not something that's
going to happen by itself. And when we come to the
question of is this technology or is it institutions--
in fact, this is a big debate in economics. Inequality-- what's the
role of institutions, what's the role of technology? And, at some level,
it's a false dichotomy. It's a false dichotomy
because I think, first of all, the two interact
in very complex ways. But most importantly-- and this
is why we put so much emphasis on what the objectives
of the corporations were, what the vision of
the tech industry is-- it is institutions, regulations,
and social norms that shape the direction of technology. But we are also insistent--
and this is partly the research that I've been doing
for many, many years-- that technology does have
a really important role in understanding inequality. So there is sort of a
left wing narrative, for example, that everything is
just about the union movement and decline in minimum wages. But the facts just
don't line up with that, and that's the reason
why I showed that chart. That chart is rather
striking because it shows just the automation
part of technology, which Simon and I think is
the most important one, really is centrally
important in explaining what's been going on
in terms of inequality in the United States. In other countries,
it's a little bit more of a complex picture because
institutions are very different and they interact
with technology. But the direction of technology,
especially the automation focus, is absolutely central. And that's why redirecting
technological change has to be quite important. Yes, indeed, railways and
steel and how they were used was critical. Again, that's the
redirection of technology. But, again, going back to
our overall interpretation, even that is not a
foregone conclusion. Companies are going to have
many different choices about how to use technologies. And the reason why
we are so insistent on digital technologies
is because they have been used in a very particular way. The conception of people like
Norbert Wiener, JCR Licklider, Douglas Engelbart, Ted
Nelson, is precisely-- our emphasis is
precisely because there was a different path of how to
use these digital technologies that was more
targeted at improving the productivity of
workers and, therefore, as a result, would have been
more beneficial for workers. But at the end,
digital technology is being used for automation,
being used for surveillance. I think those are the
choices that we have made, and those are very
consequential choices. Yes, indeed, there are
parallels from the past, but I think digital technologies
really amplify these things. And if we come to AI I
think the parallels-- again, this is why the book is
sub-entitled "1,000 Year Struggle"-- but there are also
some unique features of AI. One is the speed of change. To the contemporaries, the
introduction of textile mills would have appeared
completely revolutionary. But even then it was a slow
process over many years. For example, when you
see the weaving industry, modern factories are coming
up, but there are still thousands of handloom weavers. It takes several decades. With AI, the speed of
change is very fast, and it's also very pervasive. The potential-- again,
some of it is hype-- but the potential is
there that AI tools are going to be applied
across many industries. So that, I think, raises
the stakes, both about redirecting technological
change and the regulation. So that's part of
the reason why we think this is an
epochal time in terms of making the right choices. And it is also the reason why,
when we come to your question about the labor movement
or the general regulation, we think the focus has
to be on technology, or, at least, a very
important part of the focus. Today, we are going through
a spring of strikes. Some of it is understandable. There is a pro-union,
pro-worker president. Wages did not increase
while unemployment is low. And it is an interesting time. But with the exception of the
WGA, the Hollywood strike, I think a very, very
important absence is the discussions
about technology and how technology is
going to change things. UAW, their attitudes to
technology is just say no. I mean, if you don't come
up with a solution about how we can transition
to electric vehicles but still make that
better for workers, I don't think the
Union labor movement is going to be successful. Or let me put it another way. How many leaders of
the labor movement have really invested in
understanding AI and thinking about how it is that AI can be
used in workplaces while that's good for the labor movement? I think that's the part of it. In that sense, the WGA
was very trailblazing because they made AI as
one of the key topics, and they have the power, because
of their visibility because of their special skills,
to actually have a say. And, in fact, probably
this agreement is a success for the
Writers Guild of America. But even in that case,
I think the WGA needs to be congratulated on
making AI such a focal issue, but they did not come up
with a positive vision of how to use AI. I think, ultimately, the
labor movement-- and you might say this is
why we may need a new and refashioned labor
movement-- needs to come up with
new ideas about how to use these new technologies
that's good for workers and ultimately
acceptable for bosses. So the WGA, in my opinion,
should have articulated a vision of using AI that's
good for its members-- that's good for the
productivity and the quality of the entertainment industry-- that was different than what
the Hollywood studios wanted. And I think that would have
been much more powerful. I think that's what we need
from the labor movement. And the broader things-- the Boer book, yes, I
think Boer is 100% right. In fact, James Robinson and I
wrote an article published by, of all people, in an edited book
by Michael Lewis on the Occupy Wall Street saying at the time
that it wouldn't amount to much precisely because
it did not have a plan for institutionalizing. So the progressive
movement's success is that they wanted to change
politics, not just protest. And I think that's
what's missing. And the labor movement or a
broader democratic movement targeted at changing regulation,
changing labor market institutions, changing
technology, I think, needs to be embedded in a
bigger institutional setting. And then the final
thing I'll say about the international
dimension. I think the international
dimension is key. After all, one of
the things that's quite obvious from
this discussion is that the choices that are
being made in the United States and in China are going to
have sweeping implications for every country
around the world. But where is their voice? I think the biggest
issue is that if you want to have a direction
of technology that's responsive to global needs,
including the needs of more than 4 billion people who
don't live in Europe, China, and the United States,
then you actually need their voices to be heard. So we need international
organizations or new vehicles for the emerging
world's voice to be heard because otherwise there's
a real danger that AI is going to go in an inappropriate
direction for their needs, and, worse, it's going to
become a more and more powerful surveillance technology
that can be unleashed on the populations of
these emerging democracies. And we're already seeing that. Thank you for those
fantastic questions, Fotini. [APPLAUSE] EVAN LIEBERMAN: Daron and
Fotini, that was terrific. But now we have
almost a half hour to take some questions
from the audience. So you'll see there are
two microphones lined up on either stairwell. So if you could just line
up in those, and we will-- if we have a queue on both,
we will not discriminate against either side but rotate. And, again, if you could just
ask a question, and just one-- and I see we have someone
there, so you can start. AUDIENCE: So my name is Shahzad. I'm a grad student at Tufts. So you showed us some
data on wage inequality and how it was brought
about by task displacement and by educational
attainment levels. But, at the same
time, we all know that technological
advancement brings about huge differences, huge
improvements in life quality, right? So, for example, now we can
connect with your loved ones 1,000 miles away in an instant. Now we have AC, right? So I'm wondering, would we
have seen a different picture than the one
presented in the book if we looked at the net
effects of technology instead of focusing
just on income effects? DARON ACEMOGLU: Absolutely. I think the effects
of technology are very rich and complex. But I think the
general principle that we want to
push in the book is that for all the technologies,
including those that affect health and very different
dimensions of life, how we use them and how we
develop them is of first order influence. So, first of all, I think
some of the technologies that we are talking
about are actually making many of the goods
that we enjoy cheaper. That's already taken into
account in the numbers that I have given you,
imperfectly, albeit. But the Bureau of
Labor Statistics makes an effort to take
into account what's going on with the cost of living. So despite the fact that we
have benefited from cheap goods, the real wages of some
groups are declining. I think health is a
super interesting area. Tremendous, tremendous
advances in health technology from which, again, we have
all benefited-- antibiotics completely revolutionized
everything. Beta blockers. But, actually, the picture
is even complex even there. If you look at the US data,
it's a striking thing-- it's not COVID, so let's
stop it before COVID. From the mid-2010s to 2019, life
expectancy in the United States is falling. And is falling because low
income Americans are actually experiencing worse health
conditions, a higher mortality. So, again, despite the fact that
we have these tremendous sort of advances-- nanotechnology,
targeted medicine, and some people are spending
billions of dollars to reach 200 years of age-- people are dying sooner and
sooner in the United States. So, again, I think
there are choices that we have to make
in order to improve how those technologies actually
impact the way in which people live and benefit from them. AUDIENCE: Do you
see any potential for the development
of worker ownership in America and elsewhere? DARON ACEMOGLU: That's
a good question. That's probably much more
radical than the things that Simon and I are advocating. Absolutely, there are places
around the world where you see more
collectives, and there are a few in the United States. But I am not sure that
that's going to be-- that's got the potential to
become the model organization in the United States. And even if it
were, we would need to have more of a
discussion about the pros and cons of that. But I think the halfway house
that Simon and I are suggesting is workplaces in which
labor is treated better, it has more of a
voice, for example, through labor organizations,
there are limits to how harsh the working conditions are,
how intrusive monitoring is. I think that does not
require as radical a step. And especially if you
couple that with efforts to increase worker productivity
rather than just sideline labor, I think we're already
quite a big difference from the current environment
of how labor is being treated and where the
workplaces are going. Thank you. AUDIENCE: Hi, I'm Brian. I'm a graduate student
at the Kennedy School. So early on, you
mentioned that we need new tasks to replace
the tasks that might be displaced by automation. And my question is, what are
the characteristics that you envision for this new tasks? The reason I ask is because
I can see it going two ways. Either it becomes
more advanced tasks, which requires a
reskilling of the economy, or it's more blue
collar non-routine tasks, which would be a
decline in working conditions. So I want to hear your
thoughts about this. DARON ACEMOGLU:
Thank you very much. Yes, I wasn't sufficiently
detailed on that. Thanks for bringing it up. If you look at the
1950s and the 1960s, you see both kinds of
new tasks springing up. For example, when you can-- my colleague David
Autor and his co-authors have done that, and
Pascual Restrepo and I have done some
similar analysis as well. So you see various
measures of new tasks, and you see that both in
manufacturing, especially a lot of new blue color
tasks, and you see a lot of non-manufacturing
tasks in offices that are springing up. I mean, if you look at a-- a very crude test. If you think of
all of the people you know and the kind of
tasks that they're engaged in, actually you would
quickly convince yourself that many of them are new
relative to, say, 80 years ago. Many of the
occupations that people who will graduate from MIT--
management, consulting, design, computer programming, many
of the engineering tasks did not exist. And even for things
that existed, like being a lawyer or a
professor as an occupation, the content of the tasks that
these occupations actually perform are very different. But one issue that this sort of
reasoning immediately reveals is that, over the
last several decades, we do see new tasks,
but not enough of them. But also, the new
tasks are mostly for more college
educated workers. We see many fewer of the
blue collard new tasks, say, compared to the 50s or the 60s,
and that raises the question of, even if new tasks
come, can we then use them for creating
good jobs for workers who don't have college, which
is, of course, a very important part of shared prosperity. AUDIENCE: Thank you. AUDIENCE: Hi. My name is Callin. I'm from the Technology and
Policy Program here at MIT. I really like-- I think it's very powerful
to look at this over history and in the context of AI. But I want to try and pull out
maybe some of the differences with AI that spring out to me. And one is, with that
plot of automation against wage growth
and decline, I think something
that struck me is that one of the things that's
new about the AI discussion is the idea that a lot of the
jobs that are being automated and are now the
college-educated jobs, the ones that have
historically been growing in wages over time. I was wondering if you had
any thoughts on whether that's a fallacy, first of
all, but whether you had any thoughts on how that
changes the new tasks dimension, but also the
worker power dimension of it. Is there a dimension
of this where, because it's a new group that
is being subject to automation, the political economy of the
whole situation and the power dynamics might
shift considerably considering it's now the
more privileged groups? DARON ACEMOGLU:
Thank you very much. Those are an excellent
set of issues. They're complex ones. Let me give a two
layered answer. First of all, you're
absolutely right. There is this claim that
AI will be equalizing because it will replace or
automate more higher paid jobs. But AI or generative AI is
not the first technology that promises to do that. In the 2000s, there was a lot
of talk of digital technologies getting rid of middle
managers and so on. And what you see when you look-- this requires much more
detailed analysis-- when you look at the data with
digital technologies, first of all, you see that the
displacement has really concentrated on lower
education tasks. And the same is true
of pre-generative AI, say, the AI that was spreading
in the 2018, '19, '20. But, moreover, even when there
is some displacement of higher education workers, that's
not the end of the story because those workers can then
go and take the tasks away from lower education groups. So in the labor markets, you
actually see downward pressure on the low education wages,
even when automation targets. Some of the more
middle skill tasks. So I think my assessment--
preliminary assessment-- is that the claims
that generative AI is going to be an equalizing
force are overblown. And even if it tends to
automate higher skill tasks, that may not actually
reduce inequality. That being said, in the
book, Simon and I argue-- and Simon, David Autor, and I
have a new sort of policy paper that provides more
details on this-- there are many possibilities for
using generative AI as a tool for reducing inequality,
but it would go through, just like the
previous question was emphasizing, through new tasks. Creating, for example, finding
ways of deploying generative AI in a way that's going to be
useful for people who are more engaged in manual occupations--
electricians, line workers-- so that they can
be trained faster, they can use that
for problem solving or dealing with more real time
issues that are more complex. For example, as you get the
electrification of the grid, you're going to need much
more complex tasks being performed by electricians. So there is a path
for generative AI, but, decidedly, that's not
the one we are on right now. AUDIENCE: Hello. My question is
actually not about AI, but more about the
energy transition, where we'll also see a lot
of transformation of jobs. So currently there
are many countries around the world where the
majority of their exports and the foreign currencies
that they receive are tied to fossil fuel related
products or commodities. And what I see-- based on what I have what I see
is that these countries, most of them will not transform
in time before we will need much less of these commodities. And my question is,
what will happen to those countries and the
workers that are inside those countries-- countries
like, for example, Russia, Azerbaijan,
and Venezuela-- that are very much related
dependent on exports of fossil fuels? DARON ACEMOGLU: Yeah, those
are very good questions. And I think that
transition is not easy. In fact, that transition is
not easy in the United States. The evidence is that when there
were environmental regulations that were introduced in the
United States that helped reduce pollution, for
example, the Clean Air Act, the communities that were
focused, for example, in coal production, they were
negatively affected. And the workers did
not easily relocate to other jobs either because
there weren't the programs or because their skills
were very specialized. So that's part of the
unfortunate thing, and part of the reason why the energy
transition is so hard. I think the energy
transition is going to have distributional
effects, and some of those are going to be negative. And I think part
of the reason why this has become such a
politically charged issue in the United States is because
the pro-climate groups are perhaps not articulating
a way in which they coal miners are going to be
dealt with, how you can do this without harming the coal miners
or workers who are currently working in dirty industries. And I think that's going to be
part of the challenge moving forward. AUDIENCE: Thank you. AUDIENCE: Thank you. My question is
about how to empower people to look for better
jobs or advocate for better positions. You mentioned, for
example, unions, but I guess maybe
the people already need to have some
kind of basic safety so that they feel comfortable
to work in the union, like health insurance
if they lose their job or things like that. So do you think there's a way
to get that in, like why-- DARON ACEMOGLU: I think
that's a great question, and unfortunately I
don't know the answer. So it's very easy to agree
with what Fotini said. The three very big movements-- Occupy Wall Street, MeToo,
and Black Lives Matter-- did not coalesce into
a bigger reform agenda. It's easy to bemoan
that, but it's not easy to articulate
how they could have done that in a much more-- I think that is going to be
a very critical question, and I'm not sure that
I have the answer. But I think one
experience from history is that the labor movement has
been the linchpin of something like that because it is
much more broad based. It is about bread
and butter issues. So it creates a better
grounding for civil society and other types of
political movements. And, in some sense, perhaps
the unfortunate thing of many of these movements
that we're talking about, they have been
completely divorced from the labor movement. AUDIENCE: Hello. Hi, Professor. Thank you so much for the
sharing your thoughts so far. I'm a final year PhD
student here at MIT. And, just personally, I
came here with the intention to use technology for good. And as I plow through
the PhD, the world just looks gloomier, and, to
name a few political tensions, COVID and energy
crisis, et cetera. So going forward, I guess for
my generation for the decades to come, in your view, what
is the greatest technology fear that you have, and
what is the one thing you're most optimistic about? Thank you. DARON ACEMOGLU: Well,
that's a very hard question. I mean, I think your
attitude is exactly right. I think there is an
ethical responsibility. And the kind of questions
that you've just posed-- that technologists
need to get engaged with. I think for a long time-- it's a caricature-- but
the attitude epitomized by Facebook's motto "move fast
and break things" was, I think, defining for the
technology sector. We'll just disrupt things, and
then the social consequences don't matter. Everything's going to work out. I think we need to
move away from that. Once you move away
from that, I think there are so many different
things that can be done. But the central emphasis
that I've tried to put here is that it's not just like
technologies to cure cancer. It's not just technologies to
detect pandemics in advance. Even when it comes
to technologies that are the bread and butter
of the production process, there are huge
distributional consequences. And I think the
questions about, what are their social implications,
that needs to be factored in. And, to me, triggered
by all of these things, I think very
transformative technologies have to include
those that are going to find ways of making
workers more productive, including workers
with diverse skills, not just those with
PhDs and masters, but workers who have high school
degrees or vocational skills. How can we make workers
with every level of skills more productive in the
production process? I think that's going to be
a critical part of socially beneficial technologies
in the future. AUDIENCE: Thank you,
Professor Acemoglu, for an excellent presentation. My question will be actually
about the inequality among the countries, rather
than inside an economy. More about macroeconomic
inequality. You gave the example of
how the cotton gin actually created slave societies
inside of the United States, but the cotton gin and
the Industrial Revolution also created a huge wave of
colonization around the globe. I would like to ask how
this new wave of technology can create a new world order,
and how can this new world order change the inequality
among the countries, and what will be
your suggestions for preventive measures to
prevent any event that we have actually witnessed during
and after the Industrial Revolution? DARON ACEMOGLU: Those are very,
very good questions as well. Let me just answer it in
the context of the issues that we've talked about. The fear about AI that I
tried to briefly articulate in the one slide I put up
on the developing world is that if AI goes down the path
of more and more automation, it would actually be-- it could actually be
something that expands inequality between countries. Because if you look at the
emerging world in the last 50, 60 years, almost all
of the rapid growth experiences from South Korea,
Taiwan, Malaysia, and later China, they all leveraged
their labor force, their human resources,
which is absolutely normal because that's the
comparative advantage. Rich countries are very, very
intensive in capital goods, they have some very
educated workforces. But countries like
South Korea, China, used to have decent
workforces that are not paid very high wages. That creates a
competitive advantage. So if you push technology more
and more towards automation, you're really
precluding this type of growth behavior,
which would ultimately slow down global convergence. AUDIENCE: Thank you. EVAN LIEBERMAN:
Just a few minutes left, so if everyone could
be succint in their question. AUDIENCE: OK. My question is,
briefly, why can't we approach this AI
and optimization problem as a strictly
redistribution problem, and why should we try
to redirect technology at all to create new
tasks that we might not need in the future? And I understand the
climate technology analogy where we were able to redirect
technology to an extent, but it might also
be argued that, albeit at a horrible
environmental cost, we would have much more
energy at our disposal through fossil
fuels if we didn't have to care about those
environmental externalities. So can we use more
aggressive redistribution to handle those externalities
in the case of AI? And I know that you're critical
of UBI in the book, but maybe-- EVAN LIEBERMAN: We have a bunch
of questions on the table-- DARON ACEMOGLU: OK. Let me summarize
the point, which is, what's wrong if I makes
only 10 people around the world productive, and they
become so productive and they produce
everything else, and we redistribute
the rest to the rest? I think the problem
with that is, I think, it's both infeasible
politically and actually undesirable, even
if it were feasible. The reason why it's
politically infeasible is because if you really have
a very small group of people who do all the
contributions, they're going to become politically
very, very, very, very, very powerful. You're already not going to
convince people like Sam Altman and Elon Musk and
Mark Zuckerberg to agree to redistribution
if they become 50 times more productive
and everybody else becomes more dispensable. The political economy
of it would be worse. But even if it were
feasible, I think that would be a very dystopian
world in which social status would be highly unequal. There is a 90%,
95% of the people who are not real
contributors, and just a group of geniuses who do
everything that's worthwhile. I think that would be a
very, very unequal world. Perhaps not in terms
of consumption, but in terms of a lot of
the other things that people care about. AUDIENCE: Hello. Thank you very much
for your presentation. I'd like to make a
question about how to redirect technology
to enhance democracy, considering that
over the last years that we've seen that technology
has increased polarization and made the public
debate very problematic. So I'd like to
listen your thoughts on how the technology
could be used to make better decisions
and more constructive debates instead of
increasing polarization? DARON ACEMOGLU: So that's
a fantastic question, and the honest answer
is, I don't know. But let me say two
more sentences. First of all, I think
when the internet and then later social media
first came out, many people thought
that these were going to be democratizing
technologies. They turned out to be
wrong, but they were not completely delusional. When you look at the
structure of social media and the internet, it
creates pathways for people to interact on a
level playing field without as much hierarchy, to
form more deliberative bodies. And there are models,
both in the United States and in other countries, that are
actually attempting to do that. One that we discuss in
the book is, for example, what Taiwan has done under the
leadership of Audrey Tang that has actually led to
tangible results. Whether it can be scaled
up, whether it can actually survive the other
uses of technology, for example, for censorship
and surveillance, whether it has a future
once we have reached this level of
polarization, those are all excellent questions. But I think technically
it's feasible. And then the question is, can
we socially go in that direction as well? Thank you. AUDIENCE: Hi. My name is Nectarios,
and I basically also wanted to ask you about
the role of democratizing in AI, and to pick up one of
Fotini's questions. Do you think we need
kind of inequalities to foster progress? DARON ACEMOGLU:
Yeah, I mean, I think that's a great question
that Fotini asked, and sorry I did not
respond to that one. Yeah, I mean, I think that's
very standard in economics that I think most
economists would say, yes, it's going to be impossible
and probably quite costly if you try to create a
perfectly equal society. So that's why the
focus that we have here has been not on eliminating
inequality but keeping it to a reasonable level. I think when you have CEOs
earning 1,000 times as much as their line workers--
or even worse, the pattern that I showed, some
people are really becoming much richer and other people are
actually becoming poorer and are becoming unhealthy
and are becoming more and more marginalized-- I think those are the
kinds of alarming things. And I think those are the
ones that are reversible. No, I don't think
the dream of creating a Communist society of
everybody is equal-- that's neither feasible nor
what we're talking about. EVAN LIEBERMAN: We're
just about out of time. I'm going to ask all of you
to quickly ask your question-- DARON ACEMOGLU: And
I'll select one of them. EVAN LIEBERMAN: Any of them
that you want, or maybe none. But we'll ask you to
ask your question, then we'll get the
last two questions. AUDIENCE: Yeah,
I'll keep it short. Sorry. We've been seeing a
trend that I think people are finding college
more undesirable now, it's getting more unpopular. I was wondering if that poses
any challenges to task creation or coming up with new
tasks and automation? EVAN LIEBERMAN: You
ask your question. AUDIENCE: Hi. My question is, with
authoritarian governments utilizing new technologies
to increasingly surveil their citizens as well
as subversive powers within democratic governments
increasingly doing so as well, what steps would you suggest
that citizens in the public take to try and push back
against this from happening, and do you think
they'll be successful? EVAN LIEBERMAN:
The last question. AUDIENCE: Hi. I'm afraid of tech and
polarization and the way it impacts societies,
so do you think our polarized communities
will hinder the progress of the future to come? DARON ACEMOGLU: I
couldn't hear the last-- AUDIENCE: Do you think
our polarized communities will hinder the progress
of the future to come? DARON ACEMOGLU: That's
a very interesting-- so let me just say one
word on each one of them. I think that's a very
interesting question. I think polarization has
actually been much more micro, and inequality is much more
micro in the United States with segregation. I think that raises a lot
of interesting issues. We haven't really
dwelled on them. It's something I've
worked on in the past, but let me not say more on that. On what is it that we can do
in terms of pushing technology in a more democratic
direction and resist sort of surveillance,
again, I think that's a fantastic question. Some of the ideas
like, for example, creating more ownership of
data, collective ownership of data and better regulations
for privacy protection, I think, are part of it. But, also, I think if there
is demand from civil society and for organizations and
individuals for tools that are better at protecting their
privacy against surveillance, against censorship,
I think that could be a trigger for the development
of these kinds of technologies. And then finally on college. I think that's an
interesting-- there have been a couple of news
articles on that, that college is becoming less central. It's hard to know where we are. Actually, when you look at
data over the last five years, still the status gap, income gap
between college and non-college is as strong as before. So I'm not sure whether
this sort of college is not so important is a
reaction to a lot of people who went to low quality
colleges and they're not getting the returns, or
it's a passing fancy. We'll see. EVAN LIEBERMAN: Great. Well, I really want to
thank Daron Acemoglu. That was terrific. And also to Fotini Christia
for all of your comments, and all of you. Those were great questions. I'm delighted that MIT Center
for International Studies could host this event. Please join our
email list, which is outside, if you
want to receive updates on upcoming events. There's a sign-up
sheet in the lobby. And for those of you who
would like your book signed, I understand that Daron
is willing to do so, so please make your
way up to the lobby. Thank you again for joining us,
and please show your applause. [APPLAUSE]
