Impulse speech: Technology and banking
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
Title |
| |
Title of Series | ||
Number of Parts | 6 | |
Author | ||
Contributors | ||
License | CC Attribution 3.0 Unported: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
Identifiers | 10.5446/48112 (DOI) | |
Publisher | ||
Release Date | ||
Language |
Content Metadata
Subject Area | |
Genre |
00:00
Computer animationLecture/Conference
02:34
Lecture/Conference
03:08
Lecture/Conference
04:06
Lecture/Conference
04:58
Lecture/Conference
07:28
Lecture/Conference
09:31
Lecture/Conference
10:59
Lecture/ConferenceMeeting/Interview
15:01
Lecture/Conference
16:47
Computer animation
17:42
Lecture/Conference
19:34
Lecture/ConferenceMeeting/Interview
20:56
Diagram
22:29
Lecture/Conference
23:16
Diagram
24:52
Computer animationDiagram
26:10
Computer animation
29:42
Computer animation
31:14
Lecture/Conference
32:56
Computer animation
33:50
Lecture/Conference
35:15
Lecture/Conference
38:58
Computer animationDiagram
41:26
Lecture/ConferenceMeeting/Interview
42:21
Lecture/ConferenceMeeting/Interview
43:02
Lecture/Conference
43:36
Lecture/Conference
Transcript: English(auto-generated)
00:15
Hello, ladies and gentlemen, and welcome to this session on technology and banking. And yes, indeed, this is that dreaded post-lunchtime slot
00:23
feared by speakers and audiences alike. But in fact, you might say that we're devoting the next hour and a half to a wake-up call because in many ways, that is what technology and banking is all about. And as many of you probably know firsthand,
00:41
technology is changing Europe's banking landscape, to pick up on the title of this year's conference, in ways that many of us hardly would have thought possible just a couple of years ago. Both artificial intelligence and blockchain are revolutionising
01:00
the way that both organisations and people interact with money in a wide range of financial services and in the area of insurance. 2016 has been seen by many experts as a watershed. In fact, it's been called the year that fintech went mainstream. And if you look at the numbers, just from mid-16 to mid-17,
01:23
global investments in fintech increased 11% to a staggering 17.4 billion US dollars. Just in the second quarter of this year, over the first, they doubled. That has profound implications, both for the new guys on the block and for incumbents as well as for regulators.
01:44
And in our panel discussion later on, we want to talk about those implications. And you, ladies and gentlemen, hopefully will help us keep it lively by asking questions. Appropriately enough for a digital question, we're taking old-school questions. You can raise your hands and ask them with a microphone.
02:01
We are not using our devices on this one, just for a little bit of dialectics. So, please do keep your questions in mind. But first, I would like to take just the quickest of polls to get a sense of how you see fintech, ladies and gentlemen. So, first of all, how many of you have used some form of fintech,
02:23
whether digital and or mobile payment, cryptocurrencies, investment and or stock trading platforms or others? How many of you? Please raise your hand. Okay, very savvy audience. That's great to know. How many of you would say that the potential benefits of fintech outweigh the risks?
02:44
And when I say risks, I mean everything from business model obsolescence on the part of banks to cybercrime, abusive personal data and also systemic instability. So, how many would say benefits outweigh those risks?
03:03
Okay, how many see it the other way around? All right, we'll see who convinces whom. Maybe we should take those once again at the end. So, let us now get started and we'll start out with what is seen by many people as the driving trend in fintech innovation at the moment, namely artificial intelligence.
03:25
Large banks and small start-ups are actively exploring ways that AI could reduce costs and improve client satisfaction. We begin with an impulse speech from the author of Rise of the Robots, Technology and the Threat of a Jobless Future.
03:41
He's a futurist and entrepreneur. He has over 25 years of experience in the field of computer design and he founded a Silicon Valley-based software development firm. He's consulted with Société Générale on creation of its new Rise of the Robots Index, which focuses on firms that are playing a significant role in the AI and robotics revolution.
04:03
Please give a warm welcome to Martin Ford. Well, thank you very much. It's really a great pleasure and also a privilege to be here to talk to you.
04:22
For nearly a decade now, I've really been focused on the potential implications of technologies like robotics and artificial intelligence. And I've really come around to the view that we could very well be on the leading edge of a very substantial disruption and I think it's a disruption that is going to completely redefine the way businesses operate and compete
04:42
and it's also ultimately, I think, going to end up putting a terrific amount of stress both on society and on the economy. Now, the issue that I've really focused on is the idea that smart algorithms, machines, robots are increasingly going to substitute for human workers. They're going to take over more and more of the work that's now being done by people
05:03
and I think that ultimately, as that accelerates, it could lead to some real challenges for us. As I've been engaged in my research, I've had a chance to talk to a lot of the people that are really on the leading edge of these technologies. In other words, the people that are driving innovations in AI and in robotics.
05:20
And what I've found is that at least among that technical group of people, there is something of an emerging consensus that maybe we really are headed toward this big disruption or perhaps we're even poised to enter an entirely new era, a time when maybe things are going to operate according to fundamentally different rules than what we've really seen in the past.
05:41
On the other hand, there's a whole lot of skepticism out there about the idea that this time is somehow really different and I think that it's probably fair to say that economists as a group tend to be a bit skeptical that this time is a different argument. And the skepticism is largely based on the historical record. This concern or fear that machines might displace a lot of workers
06:05
and maybe lead to significant unemployment, that goes back at a minimum 200 years to the Luddite revolts in England. And since then, the issue where the fear has come up again and again. I'm going to guess that most of you are probably not familiar with the triple revolution report.
06:21
But this was actually a very prominent report. It was put together by a brilliant group of people. The team that wrote this report actually included two Nobel laureates. And this report was presented to the President of the United States. And it argued that the US was on the brink of complete social and economic upheaval and chaos
06:42
because industrial automation was going to put millions of people out of work. Now that report was delivered to President Lyndon Johnson in March of 1964. So that's now well over 50 years ago. And of course, we haven't really seen that, have we? What has happened is that the economy has adjusted or people have adjusted.
07:02
They found other roles. They've moved on to do different things. And we have not seen this big disruption so far. So I think it's entirely reasonable to be skeptical. But at the same time, I believe pretty firmly that we may now be at the point where the technology is finally arriving
07:21
and that this time could well turn out to be really different. So what I'd like to do is begin by giving you just one number that I think is at least suggestive that maybe something interesting or different is going on. If you go back to 1998, in the business sector of the US economy, there were, as measured by the Bureau of Labor Statistics,
07:41
194 billion hours of labor performed in total. Now fast forward 15 years to 2013, and the output of the business sector grew by over 40% or around $3.5 trillion. And during that same 15-year period, the population of the United States grew by 40 million people, which is a lot of people.
08:02
That's roughly the population of Poland, which is, of course, one of the larger countries in the EU. So essentially, over that 15-year period, we added an entire country, a pretty large country to our population. So we get to 2013, however, and it turns out that the total amount of labor
08:21
demanded in the business sector was exactly the same, still 194 billion hours of labor. Now the fact that we produced a lot more with the same amount of labor is in and of itself not that remarkable. I mean, that's really just a story of increasing productivity. But I do think the fact that there was no new demand for labor at all
08:43
in the entire business sector, even after the entry of that many people into the population and workforce, is fairly remarkable. And I think that what it sort of implies is that businesses are increasingly relying on what you might call digital leverage. They're relying on all this information technology to produce ever more output,
09:03
but they're doing it in a less labor-intensive fashion. They're doing it in a way that relies less on human labor. And as that accelerates, I think that it could lead to some real disruption. So given that I am making the argument that this time could well be different, I think it's really important to articulate what exactly is so different
09:23
about today's information technology relative to the kinds of disruptions that we've seen in the past. And I would point to three fundamental things. And the first thing, of course, is that we have seen this ongoing process of exponential acceleration. And I think that nearly all of you have probably heard of Moore's Law,
09:41
or this idea that computers roughly double in power every two years or so. But actually, the acceleration is much more broad-based than that. It extends, for example, the communications bandwidth to memory capacity. It extends, in many cases, to software, where in some areas of software we've seen an acceleration far in excess of what Moore's Law would imply for the hardware.
10:04
But the other really key thing to understand is that this acceleration has now been going on for a long time. It's been going on for decades. If you measure from the late 1950s when the first integrated circuits were fabricated, we've seen something on the order of roughly 30 doublings
10:20
in computing power since then, which is just an extraordinary number of times to double any quantity. And what it implies is that we're now at a point where we're going to see just an extraordinary amount of absolute progress. Things are moving at a very, very rapid rate. And of course, they're going to continue to accelerate from this point. So what that means is that as we look ahead to the coming years
10:42
and especially the coming decades, I think it's very likely that we're going to see things that astonish us in terms of what technology is capable of. And the reason is that things are now moving at just a very, very rapid pace and things are going to continue to accelerate. The second key point that I would point to is that machines and smart algorithms
11:04
are, at least in a limited sense, beginning to think. They're taking on true cognitive capability. And by this, I don't mean science fiction, artificial intelligence, or robots running around acting like people, or machines that can genuinely think like human beings.
11:20
What I do mean, though, is that machines are making decisions, they're solving problems, and most importantly, they're learning. In fact, if there is one technology that is, I think, really central to this and has become kind of the driving force behind all of this, it's machine learning, which is just becoming an extremely disruptive
11:40
and powerful broad-based technology, and it's essentially the idea that smart algorithms can churn through massive amounts of data and based on that, they can learn, they can figure things out. And that's becoming just extraordinarily powerful. One of the, I think, most vivid demonstrations we've seen of that recently was what Google's DeepMind division in London did with its AlphaGo system,
12:02
which you may have heard about. Now, the game of Go is an ancient game played especially in Asia, but there are two things about the game that really kind of stand out to me. One is that the game is extraordinarily complex, much more so than chess. As you're playing the game of Go, the number of configurations
12:21
that the board can be in is almost infinite. I mean, there are actually more possibilities than there are atoms in the universe. So what that means is that you're never going to be able to build a computer system that can prevail at the game of Go the way chess was approached, which is just to throw raw, brute computational power at it and have it look ahead
12:40
and optimize moves in the future. You can't do that with Go because there are simply too many possibilities. You couldn't build a computer that fast. A second thing that really stands out is that if you were to sit down and talk to one of the very best Go players in the world, this person may not be even able to articulate
13:02
what he or she is thinking as they're playing the game. Very often they will describe it as just a kind of a feeling that they get that they should make a particular move. So when you consider those two qualities, what that suggests to me is that playing the game of Go, at least at a world championship level,
13:22
really ought to be something that's safe from automation. It's something that really ought to be reserved for people. I mean, put yourself in a position of a computer programmer imagining that you're interviewing a top Go player and trying to understand what they're doing so you can somehow put that into code. You don't even know where to start because they can't even tell you exactly what they're doing.
13:42
So this ought, to my mind, to be a job or a task that's safe from automation. And yet Google was already able to build a system that not only taught itself to play the game of Go, but then very rapidly essentially became superhuman and can now beat any human being on the planet. So I really think that that should raise kind of a cautionary flag for us.
14:04
And what it suggests is that a lot of our assumptions about what kinds of jobs and tasks and roles are going to be safe for people and not subject to the impact of technology, a lot of those assumptions are sure to be challenged. I mean, this technology is going to move into areas
14:20
that we really don't expect. I think that what we can say at a minimum is that almost any kind of job or task that is on some level fundamentally routine and repetitive and predictable, in the sense that that task is encapsulated in data so that the machine learning algorithm can predict how to do that job based on what's been done in the past,
14:42
that kind of work is going to be highly susceptible to automation and it won't matter what industry or what type of work it is. But the experience with the game of Go and AlphaGo really suggests that it could go far beyond that and it's really going to challenge a lot of our assumptions about what's going to be safe.
15:01
The third thing that I think is really crucial is that information technology and increasingly artificial intelligence is really becoming a general purpose technology and what I mean by that is that it's going to be everywhere, it's going to scale across everything, it's going to impact every industry, every employment sector. There aren't any areas of the economy that are going to be unaffected by this.
15:20
A number of people have made a comparison between information technology and electricity and I think that that's quite valid. And of course you would never ask what industries are most impacted by electricity or what jobs are most impacted by electricity. I mean, everything relies on electricity and it will come to be the case that the same is true of artificial intelligence.
15:41
Everything will rely on it and so it's going to be an incredibly broad based, very disruptive impact. I think that that is quite different from some of the impacts we've seen in the past. You can think of a classic example of what happened in agriculture. I mean, it used to be that in most countries of the world,
16:00
I guess all countries, most people worked on farms and were engaged in food production. Then agricultural mechanization came along, tractors and combines and all the rest of it and millions and millions of jobs were lost so that now in advanced countries, maybe one or two percent of the workforce is working in agriculture.
16:20
And of course that did not result in massive unemployment, not in the long run, but agricultural technology was very specific and mechanical. It impacted one sector of the economy and then people moved on to other sectors. You know, they moved on to manufacturing and later on they moved to the service sector. This time we're looking at a much more broad based impact, something that's going to really scale across everything
16:41
and for that reason I really do think that we could see a different outcome. One of the things that people who are a bit skeptical of all of this will be quick to point out is that there is this ongoing process of creative destruction. So we can acknowledge that many industries that exist today are going to be upended,
17:00
maybe even destroyed by technology, but at the same time entirely new things will arise in the future and it's easy to, you know, to anticipate what some of those are going to be, things like nanotechnology and synthetic biology and so forth. These new areas, these new industries are going to arise and they will have to employ people, but there are good reasons, I think,
17:21
to believe that these new areas may not be very labour intensive. They simply might not hire that many people. So what you see here is a comparison of a very traditional industry as exemplified by General Motors and at its peak employment, GM had about 840,000 workers. Now compare that to Google
17:41
and after adjusting for inflation, Google employed less than 5% of that number of workers, but it actually earned 20% more than General Motors ever did. So this is another example of what you might call digital leverage. You know, Google has become this incredibly valuable, powerful, influential corporation, but it's done it with a workforce
18:01
that is really just a tiny sliver of what General Motors employed. It's also important to note that the people that Google hire are quite different from General Motors. I mean, GM created a lot of jobs for typical average people. Google, on the other hand, typically hires people from the very best universities in the world, people that have very, very high levels
18:21
of capability and skill. So as we look forward, the point I'm making here is that the whole economy might gradually come to look more like Google and less like General Motors. And that could mean an absolute shortfall in number of jobs, simply not enough jobs to go around, and also potentially a skill mismatch problem where, to the extent that new jobs are created,
18:41
they might not be a good match for a lot of the average people that really need a job. Here's another graph that shows a somewhat similar issue. I would also say this is about creative destruction, but now I'm thinking in terms of occupations or job titles rather than industries. And one thing that you'll hear people say very often is that,
19:01
well, lots of the jobs that exist today are going to disappear because of technology, sure, but they're going to be entirely new kinds of work created. There'll be jobs in the future that today we can't even imagine. And that's certainly true. I mean, you can already think of jobs that exist today that maybe a couple of decades ago we wouldn't have been able to anticipate at all,
19:20
things like website designers and social media marketers and data scientists. These are all relatively new kinds of work. But if you look at the bar chart on the right here, what I'm showing is that while these entirely new occupations do exist, they really are not a substantial fraction of employment. It turns out that in the United States,
19:41
about 90% of the workforce is engaged in occupations that existed 100 years ago. So it is things like driving vehicles, preparing and serving food, working in offices and banks and retail environments and warehouses and factories. These are all occupations that someone in the year 1915
20:02
would have been quite familiar with. And it still is where the bulk of our workforce is employed. And yet we can see that many of these traditional areas are going to be disrupted. Just to take one obvious example, think of the impact of self-driving cars and trucks. That's going to be millions and millions of jobs just in that one occupation.
20:21
And of course it won't be just that occupation, it will be many others. So I think there's a real question as to whether, I mean, we can stipulate certainly that new things are going to arise, new kinds of work are going to arise, but there are really going to be enough of those opportunities to absorb all of the workers that need employment, that lose jobs in these traditional areas. And secondly, is there once again
20:42
going to be that skill mismatch issue where maybe these new jobs created really require lots of skill and education and are perhaps not a good match for a lot of typical people. So let me review for you some of the economic data
21:02
taken largely from the US but also from other areas. And I think that what this suggests is that technology is already having an important impact on both the economy and the job market. What you see here is a comparison of productivity and compensation for typical workers.
21:21
And you can see that in the first half of this graph these two lines are perfectly correlated. I mean, they move in perfect lockstep. And if you pick up any economics textbook, I mean, it will tell you that this is the way things are supposed to work. The idea is that technological progress
21:43
basically improves the machines that are used in the workplace and as a result, workers become more effective, more efficient, more productive. They are able to produce more and as a result of that, they ought to become more valuable and they ought to be able to command higher wages. And so what you see is that as productivity increases,
22:01
wages or incomes move in line with productivity. And so prosperity as a result of all this progress sort of scales across the entire workforce and the entire population. So it's a very positive story for everyone. That happened in the United States right up until around 1973 and you can see that at that point these two lines
22:21
effectively decouple. And since that time, compensation or wages for most average typical non-supervisory workers in the United States has basically stagnated. And in fact, some groups of workers in the U.S. after adjusting for inflation actually earn less than they did in the 1970s.
22:41
Productivity on the other hand has continued this sort of relentless climb there and so you've got this big wedge between the two, you know, the big gap opening up between productivity and incomes. And I would say that this has already resulted in some political disruption. If you go and look in the industrial Midwest in the United States in particular,
23:01
there are a lot of people that feel they've been left behind by progress. And this graph really shows quite vividly that that is the case. I mean, what's happening here is that really all the fruits of productivity increases and of technological progress are really accruing to the people at the top
23:21
of the income distribution, to business owners, managers, executives, investors, people like this. And average workers are getting almost none of that and of course, many people are unhappy about that and I think that that is certainly one of the things that led to the rather surprising election result that we saw in November. And it's interesting if you look at this same graph for the United Kingdom,
23:42
it looks quite similar. The decoupling takes place somewhat later but basically you see the same phenomenon in terms of this gap between incomes and productivity and of course, you saw Brexit there as well. So I think that this is already leading to disruption and I think that that could very well increase in the future. I think that
24:01
one good explanation at least in part for what's happening here is that the nature of machines and technology used in the workplace is transitioning. At one point, especially during that golden age after World War II, machines were clearly tools. They were something used by workers that made workers more valuable and more effective.
24:21
But now in many cases and at least relative to many workers, machines are transitioning to become more autonomous and what that means is that rather than complementing workers and making those workers more valuable, in many cases, they're substituting for workers or de-skilling jobs and making workers in effect less valuable. And I think that
24:40
that's one of the most important factors that's causing this big gap that we see opening up between compensation and productivity. Now here's another graph which shows labour's share of national income. In other words, the percentage of national income going to labour as opposed to capital and labour. Economists used to believe that that was fairly
25:00
fixed and stable and you can see that again in that post-war period here it's sort of gyrating around 65% or so. But then it goes into gradual decline and then around the year 2000 it really goes into quite steep free fall. It's now all the way down around 58%. So this is really the story of the rise of capital
25:21
and the fall of labour in terms of their relative abilities to capture income. Now a lot of economists have looked at this graph and come up with different explanations for it. Globalisation is certainly one of the things that might have generated this, the rise of China. In the United States there has been a rightward shift
25:40
in our general economic policy for sure. One thing that we've done in the US is we've almost completely wiped out unions in the private sector. So employees now have less ability to organise and to bargain and that may certainly be part of what's happening here. There have been other studies that have associated this trend
26:02
with the growth of the financial sector. That also can result in the falling labour share. So there are a number of potential things going on here that you can point to. But here is another set of graphs and this shows basically the same thing labour share but for a number of countries across the world. And the thing that should strike you
26:21
is that these are quite different countries with very different political systems and different internal politics. If you look at Germany and Canada for example, I think these are both countries where organised labour has not been decimated. It's still quite powerful and has a role to play and yet you see the same trend in place there.
26:42
You can look at China and that's of course the country that we most associate with globalization. That's the country that we all say is stealing all the jobs and yet the decline in labour share there has been even more dramatic than in the United States. So clearly we're dealing with some sort of a global force here. There's something
27:00
happening across the entire planet that is causing this to happen in many different countries with different environments and I think that it's pretty hard to come up with an explanation other than technology that really could explain that. Here is a graph of job creation
27:21
by decade in the United States that shows the number of new jobs created in percentage terms in each decade over roughly the last century and you can see it looks almost like a declining staircase. Just about every decade has produced fewer new jobs in percentage terms so there's some kind of a structural change happening there. You can see that the last position
27:41
there at zero is the first decade of this century and there were no new jobs created at all and that is largely due to the financial crisis. But if you look at the last visible bar here what I'm showing is job creation just through 2007 in other words up to the point where the crisis began
28:00
and you can see that if you sort of prorate that for the whole decade this pattern would have held. So clearly there is something structural happening independent of the financial crisis. The economy is simply generating fewer new jobs. Here's another graph that looks almost like the mirror image of that one and this is showing what happens when we have a recession.
28:21
When we have a recession millions of jobs are lost. Eventually they come back at least a number and total employment gets back to the same level it was at at the beginning of the recession. This graph is showing for important recent recessions how many months did it take for the job market to recover
28:40
in terms of number of jobs. You can see it's taking longer and longer so this is the story of jobless recovery. It's almost as though the job market is in a sense losing its elasticity. It's losing its ability to spring back to where it was before and again I would point to technology as being a primary culprit there.
29:00
The other thing that's happening here is that when we have a recession eventually the jobs come back but actually they're not the same jobs. It turns out that when a recession hits the first jobs to really get vaporized are the good solid middle skill middle class jobs. In other words jobs that require some level of training and education but not necessarily
29:21
an extreme amount so they're accessible to most people in the population. They have decent wages decent benefits reliable hours vacations things like this so they're the kinds of jobs that most people aspire to. These jobs are disappearing right away and then in the United States when the jobs
29:40
do come back that job creation is weighted heavily toward less desirable jobs. In other words lower wage jobs in the service sector or working at McDonalds or working at Walmart these kinds of things. These are a lot less you know they don't pay as well that the hours are not necessarily reliable the benefits are not great. So what you see
30:01
is there's lots of those jobs and then a smaller number of very high skill jobs that require at a minimum a university degree. And so this has been the story of polarization or hollowing out of middle class jobs. And again it is something that has been observed not just in the United States but in many countries including many countries
30:20
here in the EU so it's a global phenomenon and once again I believe it is technology that primarily is causing those middle skill routine jobs to disappear. I think there's also a bit of a warning to us here because in the US of course we do have a very low unemployment rate but that is largely due to robust creation of these lower wage
30:42
you know service sector jobs. And there is no guaranteed it that's going to continue. I'm aware personally of three or four startup companies for example that are focused on fast food automation. And that's one of the areas where we're seeing a lot of job growth now. you know it could well be that 10 years from now
31:01
even these less desirable service sector jobs are not going to be created in large numbers and that could you know be a real challenge for us I think. So let me talk a bit now about some of the actual technologies that are driving all of this. Robots of course have been around for a long time. If you had gone
31:20
into an automotive factory here in Germany or in any advanced country back in the 1980s say you would have already seen big powerful industrial robots at work. And these are extraordinarily efficient, precise, fast, strong machines. And what all of this automation has meant is that in advanced countries
31:41
all of the truly rote repetitive assembly line type jobs in manufacturing where you literally stand there and do exactly the same thing again and again. Those jobs have already disappeared. Quite a while ago. So the jobs that have been left for people in manufacturing and warehouse type environments are those that to some extent
32:01
rely on qualities like visual perception and dexterity. That are to some extent unpredictable. And examples of that might be loading and unloading boxes at a factory. Or perhaps picking a part out of a bin and then feeding it into the next part of the production process. These are so far
32:21
jobs that machines cannot do and so they've been safe for people. But that is something that's changing quite rapidly. What you see here is an illustration of a robot built by a company called Industrial Perception which is in Silicon Valley. And it was a startup company but a few years ago Google actually went out and bought a lot of
32:41
robotics companies and this was one of them. So this technology is now part of Google or alphabet's robotics division. But what you see is in the illustration it looks just kind of like a robotic arm and at the end of that arm is a machine vision camera. And you see it approaching this stack of boxes and notice that the boxes
33:01
are not stacked precisely. Some are rotated. There are gaps between the boxes. There are different shapes and sizes. Some have got text on them. Up until very recently figuring out how to move or load or unload these boxes would have been a visual perception and dexterity challenge that only a human
33:21
being would have been able to accomplish. But this company has now built a machine that is going to eventually be able to move about one box every second. And that would compare to about one box every six seconds for a productive human worker. So you can see right there how that's going to be
33:40
very disruptive. Also it's a machine that will work continuously. It's never going to hurt itself. It's never going to file a workers compensation claim. So you can see right there that as this technology becomes more affordable and available and reliable, more and
34:00
more businesses are going to come to rely on it. And that's going to threaten the livelihoods for a lot of people. In the United States hundreds of thousands of people do this kind of work for a living. Now there's a very conventional way of thinking about this sort of thing, which I think is largely correct. And that says that well if you've got the kind of job where you're lifting heavy boxes all day
34:20
for example, you've got in many ways a great job. I mean that's the kind of job where you're very likely to injure yourself. People that do that kind of work for decades very end up disabled. Then they can't work at all. So you know if a machine is going to come along and do this work that's not something that should necessarily upset us. Maybe we should celebrate that. And of course there
34:40
has always been a solution to this problem. The solution has been that we take the worker that loses this kind of a job and we give them more training, better, do something more sophisticated. Maybe they end up using their brain instead of their back in a much more safer, comfortable environment. Maybe they end up
35:00
working in an office rather than a warehouse and so it's more positive. What I want to show you now is the reason that maybe that way of thinking about this issue though is not going to be sustainable. Automation is going to go far beyond that. And this is taken from an article that appeared in the Wall Street Journal not too long ago
35:20
and it's focused on automation. In other words, smart software that's taking on more skilled white collar work. Now what you see on the right side here is a bar chart showing the number of employees, the head count in the corporate finance department of the largest U.S. corporations.
35:40
So corporate finance of course is, you know, accounting, financial planning, accounts payable and receivable, these types of jobs. And what you see on this red bar chart is the is that in a recent 10-year period about 40 percent of those jobs have disappeared. I mean relative to the revenue of the corporation. In other words,
36:00
like number of jobs per million dollars of revenue. Nearly half those jobs have disappeared. And it's largely because of smart software which is taking on more of the, especially more of the routine work and in particular would be things like robotic process automation, RPA. And this is really, you know, beginning to have
36:21
and is most relevant to the banking industry as well. Certainly there's I think a lot of commonality behind these kinds of jobs and the things that people do inside of banks. And again, most of this is because of so far RPA which is really not something that most people would consider to be artificial intelligence at all. It's really a,
36:41
you know, a programmatic approach to doing very routine things that can be defined in a step-by-step fashion. Contrast that with what I said earlier about AlphaGo and how it was able to automate tasks that couldn't even be expressed. And you can see that we are really just in the infancy of this and it's
37:00
going to get a lot more dramatic. It's going to go far beyond the routine work that we're seeing. But there are many examples you can give of how these technologies are climbing the skills ladder and beginning to impact jobs and tasks done by people with university degrees and even education beyond that. You know, in the field of journalism for example there are smart
37:20
algorithms that can tap into a stream of data, analyze that, figure out what is the compelling story within that data and then automatically generate a new story based on that. And you can read one of those stories. They're often published by the largest media organizations. You can read one of those stories and it won't be obvious to you that it was
37:40
written by an algorithm and not by a human journalist. In the field of law there are many algorithms that are doing document review. There are very sophisticated systems that are doing contract analysis and so forth. So even routine work done by attorneys and paralegals is already being impacted. Of course on Wall Street
38:01
there's been a fairly dramatic impact. I mean it's not too long ago that trading floors used to be full of people talking on telephones. Now that's virtually all gone. Most trading is now algorithmic. These are incredibly sophisticated algorithms that can for example tap into machine readable news stories
38:21
and trade on that news within tiny fractions of a second. They can essentially battle each other in very elaborate ways. They can place decoy trades and then withdraw those trades in an attempt to sort of fake out the opposition and stuff. So they're very very sophisticated AI algorithms
38:41
that are already out there trading on Wall Street. So I think that what this leads to is a world where virtually any kind of white collar job that involves sitting in front of a computer doing some relatively routine task
39:01
whether it's the same kind of quantitative analysis or cranking out the same sort of report or presentation again and again. All of that is going to be subject to automation. An extreme example would be a radiologist. This is a medical doctor that specializes in reading and interpreting visual images medical images
39:21
looking for tumors and things like this. And this is a job that in the US at least requires an incredible amount of education. At least four years of university, four years of medical school and then five years of residency beyond that. And yet I think radiology is a job that potentially
39:42
could be fully automated in a not too distant future. So the point here is that additional education is not necessarily any protection against this. What you see here is a graph of recent earnings for university graduates in the United States and you can see that it's actually declining. So it is true that it's much better
40:00
to get that university degree than to not do that. University graduates do earn more than people that don't have that education. But even for those that do have the degree things are not necessarily fantastic. Their earnings have actually been in real terms declining. And in the United States many people are graduating from university
40:21
and are not able to really find a job that really leverages their education. A lot of people end up for example working at Starbucks. So this is I think a real challenge for us because historically really the only solution we've had to automation and the impact of technology
40:40
is more and more education and training. And this is I think suggesting that that particular solution may not be sustainable so maybe we need to begin to think outside the box. Another issue that I've really focused on a lot and I think is very, very important is that of course if in the future we do have unemployment or wages are driven down, as jobs are de-skilled
41:01
that's a big social problem for sure. But it's also a very important economic problem because the economy, the market economy needs consumers, right? We have to have people who have both the income and the confidence to go out and spend their products and services being produced. If you don't have that then you run the risk
41:21
of a kind of a demand shortfall where there isn't enough demand to really drive the economy. Then you run the risk of perhaps stagnation or maybe even potentially a kind of a downward deflationary spiral. There simply isn't enough consumer demand out there to drive things. And it's interesting that as you look around the world today
41:40
there are a number of economies where you see very low rates of growth in some areas of the world like Japan you've even seen outright deflation. In the United States you've got a very low rate of unemployment but you've also got zero inflation almost which is quite unusual historically usually those things are
42:00
inversely related. So there is I think reason to believe that this may already be having an impact. And of course you've also got economists like Larry Summers talking about secular stagnation and so forth. And I think that certainly the impact of technology may be playing a role already but the most important thing I think is that as this accelerates
42:21
and becomes more and more important it could become a bigger issue. And I think it's something that certainly all of you and policymakers need to really keep in mind that if inequality gets worse it can have some very important economic ramifications. So finally
42:42
where this all leads to is then the question of what could we potentially do about all of this? And ultimately I think that you can have a very utopian outlook about all of this and certainly where I live in Silicon Valley there are many people that view it that way. You can imagine a future where maybe we all have to work less
43:00
we have more time for leisure more time for our families and so forth. But I do think that clearly we're going to run into an income distribution problem. A lot of people are going to be left behind and in order to solve that we need to find a way to I think eventually decouple incomes from traditional forms of work
43:21
and I think that probably the best way to do that in the long run is going to be some kind of a guaranteed minimum income or a universal basic income. And this is an idea that remains I think very radical but it's also an idea that's getting a lot of attention throughout the world. There are experiments going on in Finland for example and so forth. So I think it's a very important idea for the future and I think it's almost
43:40
inevitable that we have to move in that direction although it's certainly politically and socially a tall order. So my point in talking to you about this is really to get you thinking about these issues. I think that all of this may prove to be one of the major challenges that we face in the coming decades so I think it's very important for all of you to engage in this discussion
44:02
and debate and to start thinking about this and talking to others and I hope that as we have that conversation from that will emerge some real solutions that will help us to find a way to build a future that works for everyone at every level of our society rather than just working for a few people right at the top of the income distribution.
44:22
Thank you very much.