It’s played a major part in some of the most fundamental and significant and life-changing scientific discoveries
for more than 350 years. And Royal Society scientists continue to work at that cutting edge,
and to make a difference in terms of scientific advances. As well as its scientific research,
the Royal Society has always been outward-looking. In fact, the Royal Society has had a Foreign Secretary for longer than
the British Government have had a Foreign Secretary. And a key part of that outward-lookingness
is to work at the interface between science and policy, and the impact of science on society.
The Royal Society does that in a number of interesting and important ways, and you’ll hear about some of them this evening.
As one part of that, in April 2017, so last year, about 18 months ago,
the Royal Society launched a landmark report on machine learning, a key part of artificial intelligence.
And the report looked at the enormous potential benefits of the technology over the next five to ten years,
a number of the challenges for society that may arise because of that and a key message of that report was that we, as society,
needed to be active in our stewardship, as machine learning and artificial intelligence developed.
So, one part of that call for active stewardship is this series of lectures and discussions, the You and AI series,
of which tonight’s lecture is the fifth. And it’s a great pleasure to be able to welcome you to that series.
And it’s part of our aim in the Royal Society to foster that discussion and engagement throughout society
with some of what we’re all aware will be very, very key issues for society in the years to come.
So, it’s my great pleasure to first of all welcome tonight’s lecturer, Professor Joseph E Stiglitz.
Professor Stiglitz is an economist and a professor at Columbia University in New York. His work has had enormous impact
in shaping both academic research and thinking, and critically shaping policy.
He’s served, amongst other things, as Chief Economist of the World Bank, and contributed to advisory bodies in a number of different countries.
He’s also written a series of highly-influential and popular books that have helped shape the global debates about really critical issues
such as globalisation, trade, and its impact on economics. He’s received many, many accolades for his work,
the most significant of which was the award of the Nobel Prize in 2001. We’re also very proud that he’s a foreign member of the Royal Society.
It’s my great pleasure to call on Professor Stiglitz to deliver his lecture on the future of work.
(Applause)
Well, it’s a real pleasure to be here at the Royal Society and to talk on this subject, which I think is of enormous importance.
There’s obviously a lot of anxiety about work in all the advanced countries.
It’s not just about jobs, it’s not just about wages
and the increasing polarisation of the labour force. It’s about inequality, about what would society be without work?
And it has many dimensions that I won’t be able to go into.
It will affect people of different genders differently. Different age groups will be affected differently
by some of the changes that we’re going to be confronting. And it also has very large political consequences,
and one of the main themes is that if we don’t manage these changes in technology well,
the consequences can be very averse to our society. If you look at those groups who supported our...
I don't know what to call him. The person in the White House now...
(Laughter) You realise the dangers of not managing it well,
the correlations between those who supported him
and unemployment, low wages, deindustrialisation,
the world of work, are very clear and fairly unambiguous.
So, if we don’t manage this well, it will feel not only the nativism
that we are experiencing on both sides of the Atlantic, but I think it could well feed a kind of new era of fascism
with a lot of remnants, memories, of what happened in the ‘30s.
So, I’m gonna try to describe
the challenges that we’re going to be confronting in a fairly analytic way. No data.
It’s really a conceptual approach, so how do you think about it?
And to realise that
changes in technology and our learning about the world
have really underlay the advances of our society
for over 250 years. If you go back on the data,
look at living standards, GDP per capita, whatever measure you use,
for hundreds of years, there were no changes at all. And then, suddenly, around 1750, 1800, in Western Europe,
they started to increase, and very, very rapidly. And it shows up in other data.
In the UK, people love to gather data, and we have data of the real wages of London craftsman, for instance,
going back some 800 years. And they have exactly the same pattern
of nothing happening until 1750, 1800, and then increasing.
And you get the same pattern in longevity. Life expectancy 250 years ago
was really very low. And, you know, great minds. There was just an interesting play I saw over the weekend
about the Brontë sisters. And they died at very young ages, you know, these great creative minds died under the age of 40.
And you realise what difference it makes. I mean, somebody like me, obviously, realises what differences these changes in technology have made
and all based really on the advances in science and technology.
So, in some ways, what we are seeing in AI is a continuation of those kind of advances
that have gone on for 250 years. But in other ways, they are different.
As I say, more than just a continuation of the process of advances in technology
where for a long while we constructed machines that were stronger than any humans.
And then, we constructed, but AI is more than robots which are stronger than humans.
And it’s more than the kinds of machines that can process information
much faster than any of us can process information. When I was a little kid, we used to live near a railroad track.
And the railway cars have long numbers.
And we used to be, you know, like, six, seven, eight-digit numbers. And we used to see how fast the train could go
and we would still be able to add up the eight-digit numbers as they went by.
That’s not a game anymore, because we know that computers can outpace us in recognising numbers.
And so, the ability to add up eight-digit numbers very quickly is no longer viewed as a particularly worthy skill.
But AI is also, particularly about machines that can actually learn faster than humans,
In many areas, even in quite complicated areas, machines will be able to replace humans.
About 80 years ago, there was a very distinguished Oxford economist,
Sir John Hicks, that analysed the impact of innovation
that was occurring at that time. And the language that he used was
labour-saving versus labour-augmenting innovations.
And the language I’m gonna use is very similar. We could talk about machines, artificial intelligence,
that replaces labour, human-replacing intelligence,
versus those things that actually augment our ability to do things.
And we should be familiar with both kinds. So, in the figure here, you know, it used to be thought that artificial intelligence
would be really difficult, hard, for those engaged in routine jobs.
Anything that we routinise could be computerise, robotised, and those jobs would go.
But what we now realise is that there are a lot of things that are actually fairly complicated, some even having a PhD, an MD, those jobs are gonna go, too.
And this picture here is that of a radiologist. Computers can read your MRIs and your CAT scans
better than humans can. It’s bad news for radiologists,
but the studies that have tried to break down the tasks that people do
point out that, actually, reading these x-rays or MRIs
is only a fraction, or only a part, of what people who are trained in these jobs actually do.
And so, this is an example, in part, where this will be intelligence-assisting
because this routine part, the machine will do better than a human. And that means the doctor will be free to do other things.
So, the point that I wanna emphasise is that it will have an impact on not just unskilled jobs,
but also skilled jobs. But still, the largest impacts are going to be on unskilled jobs.
And this was a task, driving a truck, that is fairly complicated.
And yet, self-driving trucks are anticipated to be dominant
within five to ten years. They keep postponing the date in which they think it’s going to happen.
So, two years ago, they told me it was only going to be five years. But now, nobody’s talking about three years from now,
it’s still five years. But it will happen.
Some of this is a political issue.
These trucks, you know, cars or self-driven cars occasionally kill people.
But, of course, so do human drivers. But when a driverless truck kills somebody,
everybody says something’s wrong with the technology. And when a human-driven car kills somebody, they don’t say,
‘We won’t allow any human to drive a car, anymore.’ So, we have a bias in favour of humans in this area.
But the fact is, the ability to respond very quickly
to the major factors that could cause an accident
is better by a computer than by a human. So, that’s an area where, again, they’ll be really labour-replacing.
The reason why this is, in part, so important is that one of the things that’s happened in the last 35 years
is that wages of unskilled workers has declined,
or at least not kept up with the pace of wages of people at the top.
So that in the United States, for instance, and this is particularly true for males,
a full-time male worker median income is the same as it was 42 years ago.
No increase in income over 42 years. And the people who have full-time jobs are the lucky ones.
They’re not the ones who are unhappy in the Midwest and the South. And at the bottom, real wages in the United States
are the same as they were 60 years ago. So, this replacing unskilled labour is going to be a serious problem.
And the largest employer of workers who are not college graduates,
you know, or high-school graduates, are truck drivers. So, and we’re talking about millions of jobs, here.
So, this is going to have significant effects, and effects about which we ought to worry.
As I said, in all of these areas, robots are actually, or machines, are better than humans.
There are problems that have been detected, for instance in facial recognition of minorities.
Their capabilities depend on data that they are given.
And what is almost as striking is that so far, in some areas, where humans have no difficulty at all,
computers have a lot of difficulty. Like sewing, where we think of that as an unskilled job,
the computers think of that as a really high-skilled job that’s beyond their capabilities.
So, there is these two strands that I’ve identified. Human-replacing machines and human-assisting machines.
And for the intermediate future, only 30-40% of the jobs are at threat.
Machines may also increase the productivity and effectiveness of humans.
In the area of science, we all know that. Most important advances in science are because of a new instrument.
New instruments allowed us to see beyond what our human abilities to see were.
And those new instruments have posed new problems that we then go about solving.
So, it’s unambiguously clear that these human-assisting innovations
are going to be important. And the balance of these two, what I call human-replacing and human-assisting machines,
will depend on the extent to which we increase skills in the labour force and some things about the nature of the evolution of technology
will be difficult for us to predict, going forward.
So, this is just by way of background. We really can’t predict precisely how many jobs are going to be replaced,
how much productivity is going to increase. But what I want to spend the next few minutes talking about
are the economic implications, including the implications more broadly for work.
As I said, overall, both of these forms of innovation increase the size of the pie.
And if you’re wondering about that picture, it took me a while,
the Royal Society provided that picture for me, and I wondered why, ‘what was that?’ And I said it was increasing the size of the pie,
so they had a baker there baking a pie. So, innovation is increasing the size of the national pie,
and that’s important. But the real debate is about how that pie is going to be divided.
So, that’s really where the debate is going to be.
So, right now, the left-hand is the current system based on standard data sets.
But what you see is the gross inequities.
53%, the blue area, is the income of the national pie that goes to the bottom 90%.
And that little slice, don’t feel too sorry for them, that gets only 9%, that 9% goes to the top one tenth of 1%.
So, what you see there is the gross inequities in our society. And almost surely, things are going to get worse in both dimensions.
That is to say the bottom is going to get a smaller share and the top is going to get a larger share.
But whether this occurs and the extent to which it occurs is really about policy.
And that’s what I’m going to talk about in a few minutes. There’s another way of looking at this
and that looks at the share of labour. And currently, the share of labour is around 62%,
a little short of two thirds. And capital share is about 21%.
And rent share, this is a term that’s coming more and more into usage,
monopoly profits, the land rents, returns that can neither be related to
investments, on the one hand, or labour on the other. It’s, sort of, the third part.
And that’s been growing, it’s now about 17%. And the future, if we don’t do something,
is going to be labour getting a considerably smaller share, capital probably won’t change very much,
but rents will go up a lot. And that’s monopoly profits and all the associated things.
So, the dismal picture that might emerge if we don’t manage it well
is that workers will be hurt, the world of work,
in two ways. Lower wages and more unemployment.
And this is just a picture of the kind of soup lines that existed in the Great Depression.
Whether the decrease in demand for labour
that will result from human-replacing innovation results in unemployment or wage decreases
will depend, to a large extent, on Government policy and how flexible labour markets are.
But in one way or another, workers are going to be hurt,
unless we put in place other policies that I’m going to describe in a minute.
So, the future is actually going to be bleak
if we don’t have the right Government policies. And the perspective that I’ve been putting forward
is that with the right Government policies in place, there can be full employment and workers can share in the future of the enhanced growth.
So, the pie is larger, there’s no reason why the workers have to be worse off.
But with the wrong Government policies, we will have
either unemployment going up or workers’ wages going down, or both.
And as I said in the beginning,
if we go those directions, it will have very severe implications for our democracy.
So, the point is really a very, very simple one. If the pie is bigger, if the size of the national pie is bigger,
in principle, everybody could be made better off. But that’s not the way the market economy on its own
naturally operates. And to see that in a fairly dramatic way,
this is a chart for the United States, a well-known chart
showing what’s happened to productivity, going back to 1948. And you see the green curve is productivity going up
and until about the 1970s, whenever productivity went up,
compensation, which includes fringe benefits and some other things, went up in proportion.
And there was a very major change in the structure in our society in the late ‘70s, the beginning of the ‘80s,
where there was a total disconnect between increases in productivity,
productivity meaning the size of the pie was growing, and what workers were getting.
They didn’t share in those increases. And this pattern is true not to the same extent,
and we don’t have as good data for Europe, but the same pattern is true in most of the European countries,
but not all. Another way of seeing this even more dramatically
is the curves in the bottom there that look like they’re the horizontal axis.
Those are not the horizontal axis. Those are the average income of the bottom 90%.
And, as you all know, economists do a lot of lying with statistics, like changing axes and things.
This is just the raw numbers as they come out. And what you see is those numbers are very low.
And you can use a microscope and you can see there’s some improvement.
But at the top, you see what’s happening to the average income of the top 1%.
And the obvious disparity between what’s happening to the average income of the bottom 90%.
We’re not talking about the people at the bottom, you know, not the poverty. This is what’s happening to the bottom 90%.
And the top is very dramatic. But the other thing I wanted to emphasise, and it comes out more dramatically in some other data
that I don’t have time to go through, here. Well, it comes out in some other data,
is that there are large differences among the advanced countries.
And this is a graph that looks, on the horizontal axis is income inequality, and the vertical axis is social mobility,
which is equality of opportunity. And what you see is that,
and income and equality is measured by a standard metric called the Gini coefficient. And social mobility
is the correlation between a parent and a child, okay? The big two points I want to get out of that figure are that
there are very big differences amongst the advanced countries
in the degree of inequality. I stress, this is the global forces that we’re talking about,
globalisation, changes in technology, are global.
And they’re affecting all the advanced countries in a similar way. And yet, how they get translated into economics is very different.
And the second one is that there are these huge differences in social mobility and opportunity.
And the other thing that should be very clear is countries with high levels of inequality, like the United States,
have low levels of equality of opportunity, of social mobility.
So, talking about the United States,
America likes to think of itself as the American dream, equality of opportunity.
But the life prospects of a young American are more dependent on the income and education of their parents
than in almost any other advanced country. So, I tell my students, there’s really one important decision you have to make in your life,
which is to choose the right parents. And if you mess up on that, the game is over.
It’s not quite that bad, but... Now, one of the very disturbing things,
which highlights one of the aspects I’ve been emphasising,
the importance of politics, is the blue bars are the level of inequality
by the standard measure of Gini coefficient in the mid-1980s.
And the orange bars are the current levels of inequality.
And what you see is that those societies like the United States, with a high level of inequality,
are also societies that have let inequality get worse. Not a surprise.
The implication of this graph and this graph is that inequality is a matter of choice,
it’s what we do, it’s our policies. And in democracies, societies, where there’s
a high level of inequality of income and wealth, that income and wealth inequality gets translated into politics.
And so, those at the top shape the rules of the game.
They shape the policies. And the result of that is you get increasing inequality.
There’s, in a way, a vicious circle, here, that societies like the Scandinavian societies
that have managed to maintain high levels of equality tend, not always, but have a force to maintain that kind of equality.
What I want to emphasise, it’s more than just a matter of redistribution. We talk about redistribution, and that’s clearly important,
but it’s also a matter of the basic rules of the game and how you structure society. So, this graph, as one example, is the difference between
the tuition fees at the undergraduate level in two parts of the UK, England and Scotland.
And, obviously, that can have an effect on equality of opportunity, going forward.
So, when we come to think of the basic rules of the game,
we have to understand that markets don’t exist in a vacuum.
If you think of the market as a game, any game has to have rules and regulations.
Markets have to be structured. And the way we structure markets affects both the efficiency of the game
and the efficiency of the economy, but also how the fruits of that economy are shared.
And this goes to every part of our legal and economic structure,
from monetary policy to corporate governance, bankruptcy law, antitrust, which affects competition, to labour law.
So, obviously, there are policies that can increase the demand for unskilled labour.
And that can increase the return to unskilled labour, that can increase the bargaining power of workers.
So, if the economy is run tight, so the unemployment rate is relatively low,
that increases the bargaining power of workers. If we have international agreements where
we have in the United States where American firms investing abroad
have better property rights than if they invest at home. That makes is more credible that a firm says,
‘If you don’t take a lower wage, I’m going to move abroad.’ That weakens workers’ bargaining power.
When we change the rules concerning unionisation or collective bargaining,
that weakens workers’ bargaining power. At the same time, if we don’t effectively enforce antitrust policies,
we have monopolies, which we do, that large percent that goes into rent was including monopoly power.
They raise prices, that lowers real wages just as much as a lower nominal wage does.
In other words, real wages is wages divided by prices. And if you increase prices, that lowers real wages.
And so, if you have a lot of monopoly power, and there’s been a large increase in monopoly power in the United States.
I haven’t studied it in the UK, but in the United States, there’s been a large increase in monopoly power, that lowers real wages and that leads to more inequality.
So, all these are mini examples of how the basic rules of the game have,
exactly at the time when they should have been done to address the problems
of deindustrialisation, labour-replacing innovation,
made things worse. And we’re just getting a picture
of how much worse things could go in the future.
So, another aspect of this is policies to sustain economy at full employment.
To give you a couple of examples, monetary policy that focuses exclusively on inflation
and doesn’t talk about unemployment, the effect of that is to lead to a higher average level of unemployment
and weakening the bargaining power of workers.
If you have fiscal policies, they unnecessarily introduce austerity,
as has happened in many countries in Europe. You have what I call deficit fetishism,
not a balanced account. Then, that depresses aggregate demand,
and that increases unemployment, and that weakens the bargaining power of workers, so workers get hurt in two ways.
And there’s a growing sense that conventional policies may not be enough.
There’s a lot of discussion beginning now in the United States of a guaranteed right to a job.
And, you know, ‘Can we afford it?’ India, much poorer than the United States or the UK,
has introduced a rural employment guarantee scheme, that guarantees 100 days of work
to the 800 million people that live in the rural sector in India. It’s not perfect, but they could afford it,
and it has had an effect of raising wages in a way that one would have anticipated it would.
So, in the United States now, there’s a growing, in a progressive...
Not Trump... but in the progressive movement, there is a growing sense
that maybe we ought to be exploring a guaranteed employment. Not necessarily at 40 hours a week,
but at some level commensurate with where we could be.
So, the bottom line is that with the right policies in place, AI can usher in a new era of shared prosperity,
with meaningful work for all of those who desire it.
Without the right policies, the kind of dystopia to which we have been moving
will only get worse. So, I just welcome you’re having this discussion on the future of work,
because I think there is no issue that is more important, not only for the economy but for our democracy.
Thank you. (Applause)
Professor Stiglitz, thank you very much for a really thought-provoking and interesting lecture. In the interest of getting to the discussion as quickly as I can,
I want to move on now to introduce our respondent, Professor Diane Coyle. Professor Coyle is the Bennett Professor of Public Policy
at the University of Cambridge. Her work in economics and the policy consequences of that
have had substantial impacts. She’s had many public service roles, and at the beginning of this year,
she was awarded a CBE for her work in economics and in the public understanding of economics. And it’s a great pleasure to invite her to address you.
(Applause)
Thank you very much. So, my job is to kick off the discussion, as if you need it,
with a few brief comments. And I want to pick up on exactly the point where Professor Stiglitz left off
and talk about policies, and make three points about the kind of policies we should be thinking about now
to deal with the prospects of what AI is going to do to the future of work.
I grew up in a Lancashire mill town in the 1960s and 1970s. And in 1978 there were 28 cotton mills around town,
and three years later there were two. And so, I know that you should never minimise the disruption costs
of these waves of technology and trade affecting people’s jobs. But the question as we were just hearing is,
is AI a technology like alarm clocks, or is it a technology like ATMs? And let me explain.
Even before my day, in those Lancashire cotton towns, there were people called knocker-uppers who used to go around with a big stick
and bang on upstairs windows to make sure that everybody got up in time to get to the mill, because if you missed getting there on time, you got fined.
And a key technology, the alarm clock, made those people completely redundant. Nobody does that job, anymore.
When ATMs were introduced, there were fears that there’d be no jobs for bank clerks anymore,
but the evidence is really clear that, actually, the number of people working as bank clerks increased steadily.
Because what happened was that the tasks that they did within their job, and the same job label, changed.
And they started doing things that were much higher value and serving customers better. So, they weren’t processing cheques,
but they were offering to help with advice about savings and pensions.
Now, we don’t know what AI is going to do, and there are a really wide range of predictions. But as we were hearing, we’re already in a situation where,
given what’s happened so far in technology affecting jobs, we have been really terrible at responding to the consequences
to the disruption to people like my aunties and uncles in the 1970s and ‘80s, and the incomes of people who are affected.
So, if we’re really rubbish in the past, and we know that we have this new big disruption coming down the pipeline,
we should be thinking right now about what to do about it. And I’m really glad this debate is now picking up.
The second point is about skills. And in previous historical industrial revolutions,
there’s been a long period of stagnant wages, increasing inequality, disruption in the jobs that people do.
And eventually, people acquired the skills that enabled them to work with the technology and not be replaced by the technology.
But that took five or six decades during the industrial revolution, and we don’t want it to take so long.
We want to know now what’s going to happen to the radiologists, and what’s going to happen to the truck drivers,
in five years’ time. And it’s been all talk and no action, on education and skills.
Politicians have been talking about education for a long time, and we still have a system where politicians think it’s really important
to memorise the plot and quotations from hard times and regurgitate them in an exam,
and there has been no real getting to grips with the education and skills and training policies we need
for the people who are going to be affected by AI. The third point, and this goes
to the last point that Professor Stiglitz was making, is that innovation is shaped as much by society as it is by technology.
And not all innovation is socially good. In the world of finance, we think ATMs were a good thing.
CDOs, collateralised debt obligations, were not a good thing. And we can see that some of the AI applications
have absolutely wonderful potential for society. The benefits they could bring in health or transportation or energy
might be really significant. We need to make sure the structures are in place for those benefits to be really widely shared.
And one of the ways that I’m most keen on is to think about competition. And, in particular, competition over access to data,
because that’s one of the bottlenecks, now. So, we might start to think about public data repositories,
and what are the conditions of access to certain kinds of databases, and what are the expectations on companies that are using public data?
We might think about those issues. We might even think about public service AI,
to change the character of competition in the market and make sure that there’s somebody who’s got a different business model
and a different set of motivations providing innovations that are there for the public good, and competing with everybody in changing the terms of competition.
So, those are a few ideas from me. And now, I’m sure that everybody here has got lots of their own. Thank you.
(Applause)
So, let me, on everyone’s behalf, thank both of the speakers for a really interesting, and as I said, thought-provoking talks.
In opening up to questions, for the people in the other rooms, remember there’s a mechanism that we’re keen you use to get your questions through,
and there’s someone sitting in the front row who will pass those on, on your behalf. Maybe I could just cheat a little by using the chair’s prerogative
and asking the first question. For those of us who are concerned, and I think it’s hard not to be concerned,
having heard both of the lectures and thought a little bit about it, what can and should we be doing as individuals, to make a difference?
Well, that’s a hard question. Let me put it in terms of, maybe, the corporate context.
What I said is that it’s how we run our society
that's going to make a great deal of difference.
And that requires spending money on education. We talked about training and retraining people.
That requires taxes. And corporations that don’t pay their fair share in taxes, it seems to me,
are not fulfilling their responsibility in helping us address this problem.
Paying their workers decent wages. You know, the model of shareholder capitalism has always been,
‘Squeeze the workers as much as possible’ because your job is to maximise your return for the shareholder.
But the societal consequences for that are very averse.
So, those are, you know, two examples. More broadly, I think the argument that I’ve been putting forward is
basically, the outcomes would depend on the rules of the game
that we set up. And therefore, it's really the political context,
it’s political engagement, and there’s a battle. I mean, I see it very strongly in the United States where
we have one party who wants the rules more tilted to the upper 1%, or one tenth of 1%,
and the other one is trying to change it the other way. So, we have a very clear battle.
So, anybody who wants to see the dystopia be brought about, they should support that other party.
But if they want to have concern about social and economic justice,
or don’t want that dystopia, there's a clear political agenda, going forward.
Thanks. Diane? So, I think I’d add, be prepared to learn new things, because I don't think anybody can relax
and assume that they’re gonna carry on doing what they’re doing now for the rest of my career. My 20-year-old is trying to teach me some new programming language,
with very limited success, on my part. But the other thing is about
thinking in terms of collective organisation. Because as individuals, there’s a limit to what we can do.
But one of the lessons of the industrial revolution was that those terms of the rules of the game started to change when people got together,
they formed philosophical societies, they formed unions, mutual, cooperatives. So, thinking about how, collectively,
we can come together and make sure that the rules of the game work for the benefit of most people.
Thank you. Let me open up to questions from the audience. Front row?
- So, I can’t help noticing that this disconnect between medium wage and productivity was in the ‘70s and late ‘70s.
And it’s about the same time that Reagan and Thatcher both were in office
and instituted policies that favoured, you know, very, sort of, strong right-wing free market policies.
With the fall of the Soviet Union and, you know, communist countries, they seem to have, sort of,
felt that they had some truth in their policies, they felt vindicated.
And what strikes me is that the Scandinavian countries today are moving more to the right.
They’re saying they can’t afford all these social benefits, they can’t afford, you know, sort of, free education, etc.
So, in a way, we seem to be losing this war of ideas.
And I wanted to know what you think of that. So, the reason for the movement towards the right
is that even in the best-performing societies, the Scandinavian countries, there is a sense of fear about their future,
including the future of work, but also the migration crisis was a trauma.
You have to understand, it can be a trauma in people’s mind, even when it’s not in reality.
And it can be a trauma that can be amplified by bad-meaning politicians.
So, for instance, in the United States, you know, you've all heard about the ‘build a wall’,
keeping out the Mexican migrants. When he started talking about that, we hadn’t had a migration from Mexico for eight years.
So, it was a problem that had disappeared. But there were a lot of people who were anxious about their job.
They displaced that anxiety, to say, you know,
‘It’s others, it’s trade, it’s immigration.’
So, my response, you know, is to say that the issue about whether we can afford it, I would say we can’t afford not to do it.
So, it’s not about what we can afford. Let me put it a slightly different way.
At the end of World War II the United States had a much higher debt-to-GDP ratio, about 137%.
Our per-capita income was a fraction of what it is today. And yet, we said to everybody who had fought in the war,
which was every young man and a lot of women, you could get as much education at the most expensive school that you can get into
for as long as you can, you know, get admitted,
and the Government will pay for it. So, if we could afford it then, we could afford it now.
And, you know, to fit in with what Diane said, that turned out to be very important in restructuring our economy
from agriculture to a manufacturing economy. That was really the critical skill transition.
So, that’s why I think... you know, I can’t answer how do I win the debate in Scandinavia,
but I can tell you the economics is very clear, it’s very clear that this is a good use of our money.
[Diane] The political trends everywhere make it all the more important that new technologies start delivering economic growth
and that those benefits turn into higher wages.
- Thank you very much. Tera Allas from McKinsey. - You’ve both emphasised skills, which makes a lot of sense.
Education is, of course, part of that, but the working-age population is also a part of that. And we know that almost 80% of the people
who are gonna be in the workforce in 2030 are already there. The UK Government spends 0.01% of GDP on workforce skills,
and it seems to be a typical market failure, where the employers think Government should pay for it,
Government thinks employers should pay for it or individuals should pay for it. Individuals probably think that it’s some combination of all of the above.
How do we square that circle? [Diane] My answer is very simple. I think Governments should pay for it, in this kind of context.
(Laughter) Yeah, Diane is right, and there’s a good economic theory behind that,
which is, in a world of mobility, companies are not going to have the incentive to provide
that kind of education, knowing that there’s a 20 to 50 to 80% chance
that they’re going to be leaving. And they want to teach you very specific skills, but not the general skills that will maximise your productivity.
Let’s take some questions from the other rooms. Trudy? Thank you. We’ve got loads of good questions,
so I’ll start with one that’s come from both rooms, actually, which is,
should we consider universal basic income as a way to face the automisation of labour?
And is it possible, without leading to inflation, that negates the effects?
Joseph, that’s something you might have thought about before. Yeah. I’m actually opposed to UBI as the solution,
and it goes back to the title of this whole discussion. It’s about the world of work.
UBI lets the Government off from the responsibility of running the economy to make sure that everybody who wants gainful employment should have a job.
I think that there is dignity to work, and that work is an important part of most people’s sense of wellbeing.
Some of my younger students tell me that that’s very 19th or 20th century,
and that they can lead a perfectly meaningful spiritual life on UBI.
(Laughter) And so, you know, I’ll leave it for you guys to decide.
But for me, work is important, and I think letting Government off
of the responsibility of creating an economic system that provides work for everybody who wants a job is a mistake.
It lets the Silicon Valley entrepreneurs off the hook as well, doesn’t it? They can say, ‘That’s an easy solution to the problem.’
So, I’m not in favour of it, either. It’s an individual solution to a problem that’s social, really, and collective.
And I prefer to think of universal basic infrastructure, which means that everybody has access to transportation,
the broadband they need, but also, the soft infrastructure like healthcare and education, and it’s the Government’s job to make sure everybody has access
to at least a minimum offer, there. [Peter] Thanks. Maybe one more question from the rooms? [Trudy] Yeah. It’s funny, the same kind of questions are coming from both rooms.
So, again, I’ll combine these. Coming from the conference room, it’s, how do you think the timeline of automation
will look within and between countries? And then, backing that up in the Cohen room, it says, should we be worried that AI services developed in China,
which is set to become the global leader, will not be translatable to the kind of data produced in the west?
[Joseph] Let me begin with the second question, because I think That is of increasing concern, the following issue.
AI depends very heavily on big data and having lots of data. And China doesn’t have the concerns about privacy that we have,
let alone Europe. And that gives them, you might say, a competitive advantage,
you know, in doing research on genes, they can ask or demand that everybody in their country spit
and collect the DNA from everybody, overnight. Very hard for us.
You know, we have to do it surreptitiously, to try to get the DNA. And we’ve been engaged in trying to do it surreptitiously,
but we still are not anywhere near what they can do.
What that highlights is that different economic and political regimes
have really difficult problems having what you might call a level playing field in international trade.
And there will be pressure, particularly in the United States,
to imitate the Chinese model. And I think it’s really important for Europe,
where there’s greater concerns about privacy, to remain firm on this and to say
it is an unlevel playing field, and we will impose trade restrictions of one form or another
to level the playing field for the unfair advantage that the lack of privacy in China gives.
I know it’s not a free trade view, but I don't think you can have perfectly free trade
in worlds with that different regulatory environment. The second part of the question,
one of the real concerns that is being raised
is the use of AI to engage in exploitation.
And, for instance, in discriminatory pricing. We’re all familiar with it in the context of the airlines,
Where as you search, they figure out who, and they’re changing the prices that you can get,
and you have to search on different computers so you can try to outsmart them. But I’m sure they’re going to figure that out, too.
But the fact is, the economic implications of this are very profound.
Because all our theorems,
all our ideas about what makes for an efficient economy, is the price system,
that everybody pays the same price, that means that the value of a good to everybody is the same. If you can engage in rampant price discrimination,
you are really undermining the basis of the efficiency of a market economy.
So, ironically, AI when it’s abused, and it is being abused,
is undermining the basis of the success of the market economy.
So, I think this is very complicated, I’m not completely sure what I think. If China develops fantastic services based on AI that it then exports,
and consumers in the US or the UK can use those then that’s terrific, and I don’t have a problem with that.
But there are, you know, as Professor Stiglitz was saying, some really complicated further questions about it.
And I don't think we entirely understand what’s going on or the extent to which those are necessarily bad things.
You know, a lot of these services are provided for free. If the radiologists are out of a job but it ends up
that I get my tumour diagnosed much faster than I otherwise would,
and the price hasn’t changed, that’s a really good thing. So, I just don’t think we know enough about how these mechanisms are playing out in the economy, yet.
Thanks. I want to go back towards the back of the hall, one with the microphone.
- Thank you very much. We currently have a 42-year low of unemployment.
- And you keep talking about how this is gonna, sort of, - have a huge impact and have structural unemployment, effectively.
- When do you think this is gonna affect us? - Because we keep saying five years, five years, five years,
- why should we change now, if it’s not really gonna affect us now? - Or is it gonna hit tomorrow?
Well, I don't think we can predict the timing. But let me say, the decrease in demand, for instance, of unskilled labour,
can manifest itself in two ways. Low wages or unemployment.
And the US labour market is a little bit, you might say, more flexible.
And what happens is the wages go down, and we have more inequality, the kind of thing that you saw there.
Other countries have provided more, you might say, wage protection,
but have gotten more unemployment. So, I think it’s already here. Really, the issue is the magnitude
and the pace with which it could get much worse. And I think there’s no way of clearly predicting how fast this is going to occur.
There have been, you know, big changes in the technology that have happened very rapidly,
but they have not actually been translated that necessarily quickly into the labour market.
And some of it will be much slower than five years, so I think we’re looking at quite an extended transition.
Yeah, I don't think we’re in a crisis, at this point, except that we are in a crisis for what we didn’t do 30 years ago.
And, I mean, you both talked about it, some of the, kind of, geographical differences. I may have understood the statistic wrong, but I heard a statistic
that in some parts of the US, the proportion of white males who have never been in work
was worryingly high. Yeah, I haven’t seen that kind of statistic, but it’s clear it’s very geographic.
And that illustrates, you might say, an important economic principle.
A lot of the standard models in economics have the idea that labour moves very freely,
there's, you know, efficient labour markets. And no labour economist buys that model,
even though many Government economists do.
Labour has all kind of rigidities, and that’s why wages in some places are much lower than in other places.
One of the reasons that manufacturing went to certain places,
say in the south in the United States, was that wages there were low.
Now, if markets were working perfectly, wages would be the same everywhere, and they wouldn’t have to go to those places.
But they went there because there were some structural rigidities in society in those places.
Now that those jobs are gone, those structural rigidities are still there. And so, with the very places
that were low wage 45 years ago and attracted manufacturing
are now being hit, but large fractions of those workers are not skilled,
they aren’t able to move, and there are many other dimensions of this we haven’t talked about,
for instance, housing. If they own their home,
the value of their house in these areas has plummeted, and the growth areas are places in the two coasts,
and housing there is very expensive, and they can’t afford to move. So, without some kind of collective action,
without some Government action, they’re almost trapped into these places
and the result of that is you get, you know, the opioid crisis,
and that makes it even less attractive for business to move there, and you get this really vicious circle.
- Actually, I have the mic.
- Shall I ask my question? Yeah, go ahead. - Yeah, so, I’d just like - to profoundly disagree with both of you about UBI.
- I think your assumption - that it’ll be provided by national Governments is incorrect. - Our own AI company
is working directly on the idea of global digital currencies with minimum basic incomes
based upon location-based economic reality. And I think that it is wrong
to dismiss your young students’ view on this, and it’s probably time to catch up with them.
[Diane] Yeah, you may be right. I’m very happy to see experiments in UBI so that we can see what emerges from them.
So, a cryptocurrency-based UBI, I’m even more sceptical about than a money-based UBI.
But I’m very happy to see people trying it. - That’s what the US dollar is.
Yeah, elsewhere I’ve expressed my scepticism of the cryptocurrencies.
You know, we have a very good currency, the dollar.
We have a less good one. Yeah, you have a less good one. But what makes for a good currency, it’s a medium for change,
a store of value, and the cryptocurrencies, values go up and down.
The only advantage that they have is crypto, and that’s not even clear that that’s...
- But that’s just the Betamax phase. [Diane] We’ll take up the discussion in five years.
[Peter] The lady over there. - Yeah, that would be really surprising - if we didn’t have a question on crypto and blockchain, today.
Actually, I just had a question from some person on LinkedIn,
they’re about to start a UBI blockchain, I’m not sure how that’s supposed to work.
In any case, thank you very much for a very thoughtful discussion, Mr Stiglitz and the Royal Society.
As far as I understand, you are not a big proponent of UBI and you think people should have a right and dignity in work.
Let’s imagine AI advances to such an extent that with all our ability and all our desire to work,
we are not able to contribute in any particular manner.
Do you think Governments should sponsor or subsidise biohacking or cyborgisation of humans?
Well, let me just say... I’ll let Diane answer that question. Okay, okay! But let me answer...
Very gracious. I do think that we may face difficulties
in getting employment for everybody and that we will, for the foreseeable future,
need some systems of social protection, like which UBI can be a part for those if we fail to do that.
So, I think we need backstops in the form of some forms of social protection.
And I’m not, you know, wedded to the idea that people necessarily have to work 40 hours a week.
We brought down the work week from 80 to 40. There’s no reason that it couldn’t get shorter, to 20 hours a week.
So, I don't think there’s any magic number, here. But I think that, as I said before,
for very large fractions of the population, it is likely that work will continue
to be an important part of their lives. And it is our collective responsibility to make sure that our economy works in ways that can deliver it.
At my age, I quite like the idea of being bio-hacked, I have to say, it’s a very intriguing question.
But I disagree with your premise, actually. I think work is continually redefined.
And work is actually what people get paid for doing, and the content of that work has changed dramatically over time.
So, the economic question is, how do we make sure there are mechanisms for humans to get paid for doing what they do, that gives them meaning?
Okay, we’re nearly finishing. We’ve got one more question from another room, and then we’ll have one more from right at the back.
This is actually quite similar to the one you originally asked, Peter, but it’s about working in a primary school,
particularly with the under-11s, and what they can be doing now to try and shore up their future.
This is from somebody who’d very much like to help them move on to being happy adults.
So, what should very young kids be doing, or how should their teachers be helping them?
I would say stop thinking about educating them in terms of content, of material that has to be stuffed in their head for them to repeat,
and start thinking about education as problem-solving techniques,
and re-arrange from a very young age the delivery for education.
Yeah, I was gonna say almost exactly the same thing, that one of the things about technology exchange is that
at our fingertips now, we have more knowledge than the largest library used to have.
So, access to knowledge, we don’t have to store as much in our brain, but we have to be able to know how to get it and how to assess it,
how to manipulate the information we’re getting, and how to be creative with it.
So, the focal point of education today is different.
And to pick up the other point that was made earlier, which is that if the pace of innovation does increase,
and I think there are, you know, reasons to believe that that may be the case, then lifelong learning becomes more important.
And that means this process that Diane described, of having to realise that
you will have to reinvent yourself more and more times, and that will be much more a part of where the education is going to be.
And maybe the third thing, to pick up the other point, is educate them on more spiritual values,
so that if there is no work, they can still be happy. Final question, the gentleman sitting at the back.
- Hi. Quite a lot of what we’ve heard - has been about things that have happened over the last 30-40 years.
- You know, whether it’s on housing, - whether it’s on the structure of economies, international dynamics.
And we’ve also acknowledged that we don’t quite know how and where AI is going to impact employment.
You know, is it going to be retirement rather than robots that are problematic for truck drivers, for example,
because of ageing demographics? So, given that uncertainty about AI and the historic biases,
can I ask Professor Stiglitz how different his lecture would have been
had AI not existed as an issue, at all?
That’s a good question. In terms of the structure, the analytics are the same.
It’s the same problem that we’ve faced with globalisation.
And the message is that AI could, and very likely will,
be presenting us with challenges, particularly in particular areas,
that are greater than we’ve had before. So, the process, you know, what we said before,
the process of technical change is continuous.
It’s not like overnight we’re going to lose 40% of our jobs, you know, that’s not what’s going to happen.
But there will be particular sectors that could be hit very rapidly with lots of jobs.
And that is something that could happen very quickly,
and in an order of magnitude faster than has happened in the past.
So, for instance, it is very conceivable to me that the conventional job of radiology
could disappear very quickly. You know, would you rather have your MRI read by a computer
that has been proven to be much better than a human?
Radiologists will still be there, but they’ll have to be redefining what their task is.
And I don't think there’s been that kind of cusp on important industries before, of this magnitude.
So, I think it is forcing us to think about some of the questions that we probably should have thought about
at the moment we started expanding globalisation 35, 40 years ago,
but we didn’t, and now, we have the advantage of hindsight to say maybe this time, we’ll get at things better.
Diane? I agree with that. But one other difference now is the character of the market for digital companies in general,
and these AI products. And that’s their what we would call winner-take-all characteristics and economics.
And I think it makes the job of working out how to have competing products and competitive markets
harder than in the case of previous technologies. The problem with monopoly,
that I talked about before, is much worse now. And the ability to engage in price discrimination
was not a problem 40 years ago. And that is a new problem. So, there a lots of little...
I mean, I wouldn't say these are little, but there are lots of details of analysis that I think have changed, now.
So, with regret, I’ll have to draw things to a close because of time. We’ve already overrun, but thank you for your forbearance with that.
Thanks to all of you for your involvement and engagement. Thanks to those in the other rooms. Thanks also to those of you who are watching online who tried to get in.
There were hundreds of people who queued and weren’t able to come in in person, and I hope you get the chance to watch it online,
and thanks for your interest. Many of the themes that have been discussed today feature in a report and evidence synthesis
that the Royal Society and the British Academy issued today on AI and work. And that’s available, it’s available from the Royal Society’s website,
for those of you who are interested in pursuing it further. As I said, this is a fifth in the You and AI series,
for which the Royal Society is very grateful for DeepMind’s support. All of the lectures up until now have largely involved presentations
and then interactions with the audience. The two final events in the series are a chance for you, the public, to ask questions.
They’re held later in the year. One of them is being held in Manchester at the Royal Exchange Theatre and one in London at the Barbican,
as I said, In a few months’ time, later in the year. The tickets for the Manchester event are now on sale,
and those for the London event will go on sale soon. Just as a bit of a heads up, the Royal Society ran an event
about eighteen months ago at the Royal Festival Hall. Again, tickets were made available at a, kind of, nominal price,
but just to encourage people who had got the tickets to come along. The week leading up to the event, the event was sold out,
those tickets were changing hands on the black market for more than tickets for Adele’s concert at the end of that week.
So, there may be a lot of demand for these events. It’s, again, a key part of the Royal Society’s vision
to engage in a debate, to encourage exactly the sorts of conversations that we’ve been having this evening.
So, I want to finish, once again thanking all of you, but to thank, very much, both of our lecturers this evening.
It’s been an extraordinary opportunity to hear your thoughts and your wisdom, in both cases, extraordinary, wise and considered views
on what is for us, I think, one of the, if not the, major challenges we have in society. So, please join with me once again in thanking both of the speakers.
(Applause)
No comments:
Post a Comment