1
00:00:07,020 --> 00:00:10,770
Good evening, very warm, welcome to the Oxford Martin School, I'm in Golden,

2
00:00:10,770 --> 00:00:18,180
I was the founding director of the school and it's a huge pleasure this evening to welcome Daniel Susskind,

3
00:00:18,180 --> 00:00:23,100
who's going to introduce us to his wonderful new book, A World Without Work.

4
00:00:23,100 --> 00:00:28,150
I'm sure you've picked up in the media the many great things that have been said in

5
00:00:28,150 --> 00:00:32,400
the very short period of time since this was released a week ago about the book,

6
00:00:32,400 --> 00:00:38,670
including that it is this The New York Times required reading for every presidential candidate.

7
00:00:38,670 --> 00:00:42,870
So first required reading for presidential candidates certainly is for all of you,

8
00:00:42,870 --> 00:00:47,910
and there will be book sales and a drink which you are invited to afterwards.

9
00:00:47,910 --> 00:00:48,630
As you may know,

10
00:00:48,630 --> 00:00:56,100
the Oxford Martin School was established to deal with the great challenges of the 21st century and to apply interdisciplinary thinking,

11
00:00:56,100 --> 00:01:00,450
including this question of the intersection of technology and society.

12
00:01:00,450 --> 00:01:08,820
And we started the group with our friend Michael Osborne in 2012, which does exactly that.

13
00:01:08,820 --> 00:01:18,240
We were somewhat criticised at times by being what people thought was rather pessimistic, saying that maybe half of US jobs were vulnerable.

14
00:01:18,240 --> 00:01:24,030
And one of the many, many things I'm delighted about was Daniel's book is that it places that in a perspective,

15
00:01:24,030 --> 00:01:29,370
it makes it seem not only prescient, but rather mainstream that work.

16
00:01:29,370 --> 00:01:38,610
Daniels, a Oxford burial person, so am I, and I just see his doctoral supervisor, David Vines, walking in thanks to such excellent supervision,

17
00:01:38,610 --> 00:01:47,430
David, which is a background to this book in one dimension, as Daniel will no doubt share with us.

18
00:01:47,430 --> 00:01:53,700
He did economics and management here and then went on to do his doctorate.

19
00:01:53,700 --> 00:01:59,490
He also worked in Downing Street in the Prime Minister's Advisory Unit.

20
00:01:59,490 --> 00:02:05,310
Not for this Prime Minister, I should hasten to add, but a previous prime minister.

21
00:02:05,310 --> 00:02:15,840
So he's been able to bridge that gap well. And what you will find in this book is absolutely remarkable ability to condense

22
00:02:15,840 --> 00:02:21,700
extremely complex economic ideas in a language which is attractive to all readers.

23
00:02:21,700 --> 00:02:29,550
It's a wonderful read, and it makes sense. And which can't be said of all my economics colleagues, and I commend it to you.

24
00:02:29,550 --> 00:02:33,300
So Daniel will speak for about 45 50 minutes.

25
00:02:33,300 --> 00:02:37,410
We'll have a short time for Q&A and then we'll have the book signing and drinks.

26
00:02:37,410 --> 00:02:54,140
The floor is yours.

27
00:02:54,140 --> 00:03:04,550
Thank you very much for that warm introduction, and it's a great pleasure to be with you this evening to talk to you about a world without work.

28
00:03:04,550 --> 00:03:09,950
And what I want to do is begin with the story, the great man your crisis of the 1890s.

29
00:03:09,950 --> 00:03:18,500
And it should have come as no surprise for some time in big cities like London and New York, the most popular forms of transport.

30
00:03:18,500 --> 00:03:27,680
The end of the 19th century have relied upon horses, hundreds of thousands of them to heave a whole variety of vehicles through the streets.

31
00:03:27,680 --> 00:03:31,970
And with these horses, of course, came manure and lots of it.

32
00:03:31,970 --> 00:03:36,260
One enthusiastic health officer in Rochester, New York,

33
00:03:36,260 --> 00:03:46,310
calculated that the horses in his city alone created enough manure to cover an acre of land to a height of one hundred and seventy five feet,

34
00:03:46,310 --> 00:03:49,130
almost as high as the leaning tower of Pisa.

35
00:03:49,130 --> 00:03:56,300
Apocryphal people at the time extrapolated from these calculations to an inescapably manure filled future.

36
00:03:56,300 --> 00:04:02,630
A New York commentator who predicted the piles would soon reach the height of third story windows.

37
00:04:02,630 --> 00:04:05,630
A London reporter who imagined that by the middle of the night,

38
00:04:05,630 --> 00:04:11,660
by the middle of the 20th century, the streets would be buried under nine feet of the stuff.

39
00:04:11,660 --> 00:04:16,760
And it said the policy makers didn't know what to do there. They couldn't simply ban horses.

40
00:04:16,760 --> 00:04:24,170
They were far too economically important. But the twist in the tale, of course, is that in the end, policymakers didn't need to worry.

41
00:04:24,170 --> 00:04:32,300
In the 1870s, the first internal combustion engine was built. In the 1880s, it was installed in the first automobile, and only a few decades later,

42
00:04:32,300 --> 00:04:36,920
Henry Ford brought cars to the mass market with his famous Model T.

43
00:04:36,920 --> 00:04:47,180
By 1912, New York had more cars than horses. Five years after that, the last horse drawn Horse-Drawn tram was decommissioned in the city.

44
00:04:47,180 --> 00:04:50,870
The great manure crisis was over.

45
00:04:50,870 --> 00:04:59,330
The parable of horse [INAUDIBLE], as Elizabeth Kolbert called it in The New Yorker, has been told many times over the years and in most tellings,

46
00:04:59,330 --> 00:05:05,660
the decline of horses is cast in an optimistic light as a sort of tale of technological

47
00:05:05,660 --> 00:05:13,400
triumph before resiliency of the Russian-American economist who won the Nobel prise in 1973.

48
00:05:13,400 --> 00:05:17,300
The same event suggested a far more unsettling conclusion.

49
00:05:17,300 --> 00:05:22,580
What he saw instead was how a new technology the combustion engine had taken a creature that

50
00:05:22,580 --> 00:05:29,210
for millennia had sat at the centre of economic life and banished it to the sidelines.

51
00:05:29,210 --> 00:05:32,450
Then, in a set of articles written in the early 1980s,

52
00:05:32,450 --> 00:05:39,890
the to made one of the most infamous claims in modern economic thought well, technological progress had done to horses.

53
00:05:39,890 --> 00:05:47,750
He said it would eventually do to human beings as well drive us out of work what cars and tractors were to them.

54
00:05:47,750 --> 00:05:55,160
Computers and robots would be to us. Today, the world is gripped again by Deonte of fear.

55
00:05:55,160 --> 00:06:02,870
Thirty percent of U.S. workers now believe that jobs are likely to be replaced by robots and computers in their lifetime.

56
00:06:02,870 --> 00:06:08,010
A similar proportion in the UK think it could happen in the next 20 years.

57
00:06:08,010 --> 00:06:15,800
And what I want to do this evening, drawing on my new book A World Without Work, is explain why we have to take these sorts of fears seriously.

58
00:06:15,800 --> 00:06:19,160
Not always their substance, as we shall see.

59
00:06:19,160 --> 00:06:25,170
But I do think some of the spirit will there be enough well-paid work for everyone to do in the 21st century?

60
00:06:25,170 --> 00:06:30,560
I think this is one of the great questions of our time, and in our in the book,

61
00:06:30,560 --> 00:06:38,750
I argue that the answer to that question is no, that the threat of technological unemployment is now real.

62
00:06:38,750 --> 00:06:44,330
Now, I don't think there's going to be some big technological Big Bang in the next few decades,

63
00:06:44,330 --> 00:06:48,080
after which lots of people wake up and find themselves without work.

64
00:06:48,080 --> 00:06:53,900
I don't think that's going to happen at all. But what I do think is that as we move through the 21st century,

65
00:06:53,900 --> 00:06:59,840
more and more people might find going to find that they're unable to make the sorts

66
00:06:59,840 --> 00:07:03,560
of economic contributions that they might have hoped to make in the 20th century.

67
00:07:03,560 --> 00:07:08,680
And I think this is a fundamental challenge to the way that we live together in society today,

68
00:07:08,680 --> 00:07:13,730
and I want to explore this problem with you this evening.

69
00:07:13,730 --> 00:07:16,880
And in particular, I want to do six things. The first is,

70
00:07:16,880 --> 00:07:21,080
I want to say a little bit about the history of technology and what I then want to

71
00:07:21,080 --> 00:07:25,460
share my general thoughts and technology and then explore one technology in particular,

72
00:07:25,460 --> 00:07:31,670
which is artificial intelligence. I want to say what I think this means for work in the future.

73
00:07:31,670 --> 00:07:37,610
I want to set out what I think the problems that this will create for us are going to be.

74
00:07:37,610 --> 00:07:46,130
And then finally, I want to end on a note of optimism, explain why, in spite of all of this, I still remain optimistic about the future.

75
00:07:46,130 --> 00:07:52,860
So first, the history, I mean, economic growth is actually a very recent phenomenon for most of the three hundred thousand years.

76
00:07:52,860 --> 00:07:57,780
Is that human beings have been around economic life has been relatively stagnant,

77
00:07:57,780 --> 00:08:02,700
but over the last few hundred years that economic activity came to an explosive end.

78
00:08:02,700 --> 00:08:09,960
The amount each person produced increased about 13 fold and world output rocketed by nearly 300 fold.

79
00:08:09,960 --> 00:08:14,280
So if you imagine that the sum of human existence where an hour long,

80
00:08:14,280 --> 00:08:20,370
then all of this actually happened in the last half, second or so in the literal blink of an eye.

81
00:08:20,370 --> 00:08:22,740
And it was Britain, of course, that led the economic charge,

82
00:08:22,740 --> 00:08:28,590
thundering ahead of others in the industrial revolution in the 1760s over the following decades.

83
00:08:28,590 --> 00:08:34,230
New machines were invented and put to use the greatly improved the way the goods were produced,

84
00:08:34,230 --> 00:08:43,650
and these new technologies allowed manufacturers to produce to operate more productively than ever before, to produce far more with far less.

85
00:08:43,650 --> 00:08:53,160
And it's here at the beginning of modern economic growth that we can also detect the origins of automation anxiety even back then,

86
00:08:53,160 --> 00:08:55,380
when economic growth was just taking off.

87
00:08:55,380 --> 00:09:05,040
People started to worry that using these machines to make more would mean less demand for their own work as well.

88
00:09:05,040 --> 00:09:13,830
The anxiety that automation would destroy jobs spilt into a protest and dissent during the industrial revolution, as is widely known.

89
00:09:13,830 --> 00:09:17,580
Technological vandalism by the so-called Luddites was rife.

90
00:09:17,580 --> 00:09:25,890
In fact, in 1812, the British parliament felt felt forced to pass the destruction of stocking frames etc. Act,

91
00:09:25,890 --> 00:09:32,480
which made the destruction of destroying machines a crime punishable by death.

92
00:09:32,480 --> 00:09:39,960
In fact, the next year, several people were tried and executed for smashing these machines.

93
00:09:39,960 --> 00:09:45,180
Importantly, though, this automation anxiety wasn't just confined to the 17th and 18th century.

94
00:09:45,180 --> 00:09:48,510
It continued all the way through right up until the present day.

95
00:09:48,510 --> 00:09:55,740
And in the last few years, we've seen a frenzy of writing and commentary reports on the threat of automation.

96
00:09:55,740 --> 00:10:03,690
But what's interesting is that if you go back to 1940, for instance, the New York Times or the debates about technological unemployment was already so

97
00:10:03,690 --> 00:10:08,790
commonplace that the New York Times felt comfortable calling it an old argument.

98
00:10:08,790 --> 00:10:16,200
In fact, going through the archives and I managed to find in every decade, almost every decade since 1920,

99
00:10:16,200 --> 00:10:23,650
an article in The New York Times engaging in some way with the threat of technological unemployment.

100
00:10:23,650 --> 00:10:26,980
And yet, and this is the key key point, really.

101
00:10:26,980 --> 00:10:34,430
And yet most of those anxieties about the economic harm caused by new technology turned out to be misplaced.

102
00:10:34,430 --> 00:10:36,940
And looking back over the last few hundred years,

103
00:10:36,940 --> 00:10:45,070
there's little evidence to support the primary fear the technological progress would create large pools of permanently unemployed workers.

104
00:10:45,070 --> 00:10:53,230
Know it's true the workers have been displaced by new technologies, but eventually most have also found new work to do.

105
00:10:53,230 --> 00:11:00,850
And so the question is why? How can it be that in the past, despite the fears of so many people, though,

106
00:11:00,850 --> 00:11:08,920
that the technological progress turned out not to lead to the sorts of levels of unemployment that they worried about?

107
00:11:08,920 --> 00:11:12,970
And the answer to that question, I think, is that when we look back over the last few hundred years,

108
00:11:12,970 --> 00:11:17,860
what we see is the harmful effect of technological change on work.

109
00:11:17,860 --> 00:11:23,230
The one that really preoccupied our anxious ancestors is only half the story.

110
00:11:23,230 --> 00:11:30,880
You know, yes, machines took the place of human beings in performing certain tasks, but they didn't just substitute for people.

111
00:11:30,880 --> 00:11:38,620
They also complemented them and other tasks that had not yet been automated, raising the demand for people to do those activities instead.

112
00:11:38,620 --> 00:11:46,180
And this helpful force so often forgot about the second force worked in a variety of different ways.

113
00:11:46,180 --> 00:11:48,610
I mean, perhaps the most obvious way that it worked,

114
00:11:48,610 --> 00:11:57,340
the first way that this complementing force has helped human beings is by making people more productive, more efficient at certain tasks.

115
00:11:57,340 --> 00:11:59,590
So a taxi driver, for example,

116
00:11:59,590 --> 00:12:07,540
can use a sat nav system to navigate unfamiliar roads or an architect can use computer assisted design software to design bigger,

117
00:12:07,540 --> 00:12:17,380
more complex buildings. But economic history also reveals a second way less direct way that this complementing force has helped human workers.

118
00:12:17,380 --> 00:12:24,070
If we think of the economy as a pie, technological progress over time has made the pie far bigger.

119
00:12:24,070 --> 00:12:28,300
As productivity increases, incomes rise and demand grows.

120
00:12:28,300 --> 00:12:35,040
So the British pie, for instance, is more than 100 times the size it was 300 years ago.

121
00:12:35,040 --> 00:12:38,250
And intuitively, it's possible to see how this might have helped displaced workers,

122
00:12:38,250 --> 00:12:47,610
people displaced from tasks in the old pie could tumble into task relativities in this new, bigger, better pie instead.

123
00:12:47,610 --> 00:12:52,290
And in turn, there's also a third way that the complementing force has health workers.

124
00:12:52,290 --> 00:12:59,850
Technological progress is not only made the pie bigger, but has also changed the pie to again think of the British economy.

125
00:12:59,850 --> 00:13:05,760
Not only is it for more than a hundred times larger than it was 300 years ago,

126
00:13:05,760 --> 00:13:10,240
but that output and the way that help is produced has completely transformed by.

127
00:13:10,240 --> 00:13:18,300
Five hundred years ago, the British economy was made up of farms. Three hundred years ago, factories and today of offices.

128
00:13:18,300 --> 00:13:24,330
And again, we can see how this might have helped displaced workers in the past as the economy changes over time.

129
00:13:24,330 --> 00:13:34,950
People displaced from tasks in the old pie could again tumble into performing different activities and tasks in this new changed bit of pie instead.

130
00:13:34,950 --> 00:13:38,550
So distinguishing clearly between this substituting force in this complementing

131
00:13:38,550 --> 00:13:44,520
force helps to explain why anxieties about automation were repeatedly misplaced.

132
00:13:44,520 --> 00:13:52,500
In this clash between these two fundamental forces, our ancestors tended to pick the wrong winner, and time and again,

133
00:13:52,500 --> 00:13:56,700
they either neglected the complimenting force altogether and just ignored it,

134
00:13:56,700 --> 00:14:01,980
or they mistakenly imagined that it would be overwhelmed by the substituting force.

135
00:14:01,980 --> 00:14:07,530
And that is why they repeatedly underestimated the demand for the work of human beings that would remain.

136
00:14:07,530 --> 00:14:13,830
There has always been by and large enough to keep everyone in employment.

137
00:14:13,830 --> 00:14:23,910
So I want to say a little now about technology. You know, every day we hear of systems and machines taking on tasks and activities.

138
00:14:23,910 --> 00:14:32,700
Until recently, we thought only human beings alone could ever do draughting legal contracts and making medical diagnoses,

139
00:14:32,700 --> 00:14:39,060
designing beautiful buildings and composing music, writing news reports.

140
00:14:39,060 --> 00:14:45,030
Given the time constraints today, I just want to give you a general sense of how it is that I think about technology and the way

141
00:14:45,030 --> 00:14:49,740
in which I think it's useful to think about technology when thinking about the future of work.

142
00:14:49,740 --> 00:14:58,300
So although machines can clearly do more today than they can than they could in the past, they cannot do everything right.

143
00:14:58,300 --> 00:15:02,850
You know, there are still limits to the harmful substituting force.

144
00:15:02,850 --> 00:15:07,920
The problem is that those boundaries are really very unclear and they're always changing.

145
00:15:07,920 --> 00:15:14,280
And lots of recent articles and books and reports have tried to work out the new limits to machine capabilities,

146
00:15:14,280 --> 00:15:17,160
and they've done this using a variety of different approaches.

147
00:15:17,160 --> 00:15:23,520
So one approach has been to try and identify which particular human faculties are hard to automate.

148
00:15:23,520 --> 00:15:30,990
So a popular finding, for instance, is that new technologies really struggle to perform tasks that require social intelligence,

149
00:15:30,990 --> 00:15:34,500
things like face to face interaction or empathetic support.

150
00:15:34,500 --> 00:15:45,690
So from 1980 to 2012, jobs that require a high level of human interaction grew by about 12 percent as a share of the US workforce.

151
00:15:45,690 --> 00:15:51,960
A different tack has been to consider not faculties the human body human beings bring to bear,

152
00:15:51,960 --> 00:15:59,280
but instead to consider the tasks themselves and ask whether they have features that make them easier or harder for a machine to handle.

153
00:15:59,280 --> 00:16:03,810
So, for instance, if you come across a task where it's easy to define the goal,

154
00:16:03,810 --> 00:16:09,480
straightforward to tell whether or not that goal has been achieved and there are lots of data for the machine to learn from,

155
00:16:09,480 --> 00:16:16,230
then chances are that task can probably be automated. The classic example of this is identifying cats, and the goal is simple.

156
00:16:16,230 --> 00:16:21,870
Just answer the question Is this a cat? It's easy to tell whether or not the system has succeeded.

157
00:16:21,870 --> 00:16:27,450
Yes, this is indeed a cat. And there's lots of photos of cats out there on the internet.

158
00:16:27,450 --> 00:16:35,190
Perhaps disturbing things. Disturbingly so six point five billion photos of cats supposedly had access them.

159
00:16:35,190 --> 00:16:41,010
So you can imagine the other tasks which are perhaps harder to define the goal, how to see the goals being achieved.

160
00:16:41,010 --> 00:16:46,080
Well, there's not a lot, not a lot of data to learn from are harder to automate.

161
00:16:46,080 --> 00:16:51,690
The obvious problem, though, with trying to identify the limits of machines in either of these ways is that

162
00:16:51,690 --> 00:16:55,680
any conclusions you reach are going to become outdated incredibly quickly.

163
00:16:55,680 --> 00:17:04,710
Those who try to identify the boundaries of machine capabilities are like the sort of proverbial painters of the Forth

164
00:17:04,710 --> 00:17:11,850
Rail Bridge in Scotland Yard Bridge so long that those painters supposedly had to start again when they got to the end.

165
00:17:11,850 --> 00:17:13,450
Because by the time they got to the end,

166
00:17:13,450 --> 00:17:21,420
the paint and starts to peel and spend time coming up with a sensible account of what it is that machines could do today.

167
00:17:21,420 --> 00:17:26,130
And by the time you finish, you're probably going to have to start again and readjust.

168
00:17:26,130 --> 00:17:35,090
So the argument that I make is that a better way to think about technology is to stop trying to identify specific limits and repress that temptation.

169
00:17:35,090 --> 00:17:40,370
Taxonomies and instead trying to make out the more general trends, and I think when you do that,

170
00:17:40,370 --> 00:17:46,700
what you see beneath the particular ripples of progress that we see around us today are some deeper currency,

171
00:17:46,700 --> 00:17:53,330
although it's very difficult to say exactly what it is the machines will be capable of doing in the future,

172
00:17:53,330 --> 00:17:58,130
it's pretty certain that they're going to be able to do more than they can at the moment.

173
00:17:58,130 --> 00:18:06,780
Over time, machines will gradually but relentlessly advance further and further into the realm of tasks performed by human beings.

174
00:18:06,780 --> 00:18:13,640
You know, take any technology that exists today. Open your laptop, pick up your smartphone and you can be pretty confident.

175
00:18:13,640 --> 00:18:19,340
That's the least advanced it's ever going to be. And I call this process task encroachment.

176
00:18:19,340 --> 00:18:24,440
And I think when you look at the three main capabilities that human beings draw on in their work,

177
00:18:24,440 --> 00:18:28,340
whether it's manual capabilities that involve dealing with the physical world,

178
00:18:28,340 --> 00:18:36,120
cognitive capabilities that draw on our ability to think and reason or effective capabilities, our capacities for feelings and emotions.

179
00:18:36,120 --> 00:18:43,850
I think what you see in each of those areas are machines gradually encroaching on more and more tasks that require those faculties.

180
00:18:43,850 --> 00:18:52,100
Each of those capabilities from human beings. And if you have a look at the book, what you'll see are hundreds of examples of this at work.

181
00:18:52,100 --> 00:18:57,740
I think it's important, though, to say that the examples I give in the book are not meant to be exhaustive.

182
00:18:57,740 --> 00:19:04,670
There's some impressive ones almost certainly missing in a few years time when we look back at what I wrote.

183
00:19:04,670 --> 00:19:13,190
Some will almost certainly look tired and out of date and the claims as well of the companies I talk about aren't meant to be taken as gospel either.

184
00:19:13,190 --> 00:19:19,280
At times, I think it can be hard to distinguish serious corporate ambitions and achievements

185
00:19:19,280 --> 00:19:26,150
from provocations drawn up by marketeers whose job it is to exaggerate for a living.

186
00:19:26,150 --> 00:19:33,230
The icing on the cake for me was when someone asked me if I would like an artificially intelligent toothbrush for Christmas.

187
00:19:33,230 --> 00:19:37,130
I'm quite sure how intelligent you need to be to brush your teeth.

188
00:19:37,130 --> 00:19:44,690
The point is this, though, to dwell for too long, and it's just a general point to dwell for too long on any particular omission or exaggeration,

189
00:19:44,690 --> 00:19:47,630
I think, is to miss the bigger picture of what's going on.

190
00:19:47,630 --> 00:19:55,660
Machines gradually encroaching on more and more tasks and activities that in the past have required a rich range of human capabilities.

191
00:19:55,660 --> 00:20:01,550
Now, economists are pretty reluctant to label any empirical regularity as a law or a rule.

192
00:20:01,550 --> 00:20:09,770
But this process of task encroachment, I think, has proven to be as more like a zany historical phenomenon can be and barring catastrophe.

193
00:20:09,770 --> 00:20:12,140
I think it's pretty certain to continue,

194
00:20:12,140 --> 00:20:19,850
and I think this is how we ought to think about technology in general in terms of this process of task encroachment.

195
00:20:19,850 --> 00:20:24,650
What I want to talk a little bit more about now. There is one technology in particular, which is artificial intelligence,

196
00:20:24,650 --> 00:20:30,740
and it's something that has really captured people's imaginations in the last few years is something responsible,

197
00:20:30,740 --> 00:20:35,870
I think, for the renewal of worries about what technology is going to do to work.

198
00:20:35,870 --> 00:20:41,180
And I think something significant has happened in the field and I want to share that with you now.

199
00:20:41,180 --> 00:20:45,620
And the story I want to tell begins in the first wave of artificial intelligence.

200
00:20:45,620 --> 00:20:47,960
The first wave of artificial intelligence.

201
00:20:47,960 --> 00:20:55,790
This was when my dad, who was my co-author on the previous book, that I wrote The Future of the Professions.

202
00:20:55,790 --> 00:21:00,740
My dad was doing his doctorate here on artificial intelligence in the law.

203
00:21:00,740 --> 00:21:05,810
So he was back then already trying to build systems that could solve legal problems.

204
00:21:05,810 --> 00:21:14,900
And something quite interesting happened in 1986, which was a very difficult piece of legislation was passed in the UK called the Latent Damage Act.

205
00:21:14,900 --> 00:21:20,990
And it turned out that the leading experts in the world at the time on this piece of legislation was a man called Philip Carter,

206
00:21:20,990 --> 00:21:25,280
and Philip happened to be the dean of the law school here,

207
00:21:25,280 --> 00:21:29,150
where my dad was doing his doctorate and Philip came to my dad and said, Look, this is absurd.

208
00:21:29,150 --> 00:21:35,190
Any time anyone wants to understand if this legislation applies to them, they have to come to me.

209
00:21:35,190 --> 00:21:41,840
You know, why don't we instead join forces, work together and build a system based on my expertise for other people to use it?

210
00:21:41,840 --> 00:21:53,240
And that's exactly what they did. From 1986 to 1988, it was known as an expert system expert system because it was based on Philip's expertise.

211
00:21:53,240 --> 00:22:00,380
This was the home screen designed for that system. My dad assures me this was a cool screen design in the 1980s.

212
00:22:00,380 --> 00:22:06,050
Ever been convinced of that? Just to give you a flavour of what it was, they were up against the extract from the legislation.

213
00:22:06,050 --> 00:22:10,910
Section two of this act shall not apply to an action to which the section applies.

214
00:22:10,910 --> 00:22:14,660
And that's a more readily understandable piece of the law. And I love this.

215
00:22:14,660 --> 00:22:20,390
They published it in the form of two floppy disks time with floppy disk was caused genuinely were floppy,

216
00:22:20,390 --> 00:22:25,100
and essentially what they did together was they built a gigantic decision tree where you answered yes or

217
00:22:25,100 --> 00:22:30,920
no questions and you navigated through this huge tree that literally had millions of branches through it.

218
00:22:30,920 --> 00:22:37,370
And my dad and his colleagues had manually painstakingly crossed it, and they weren't just doing this at all.

219
00:22:37,370 --> 00:22:43,340
This was the general approach in artificial intelligence. And in the beginning, like my dad,

220
00:22:43,340 --> 00:22:48,320
most people in artificial intelligence thought that building a machine to perform

221
00:22:48,320 --> 00:22:54,170
a given task meant observing how human beings perform that task and copying them.

222
00:22:54,170 --> 00:22:59,220
So some people tried to replicate the actual structure of the human brain.

223
00:22:59,220 --> 00:23:05,640
Others tried a more sort of psychological approach and tried to replicate the thinking, process,

224
00:23:05,640 --> 00:23:11,670
thinking and reasoning processes in which human beings appeared to be engaged when they were performing a task.

225
00:23:11,670 --> 00:23:18,870
And the third, and this is what my dad was doing in building those expert systems, was trying to identify the rules the human beings follow.

226
00:23:18,870 --> 00:23:23,760
But all of these efforts in this first wave of artificial intelligence in all of them

227
00:23:23,760 --> 00:23:31,310
human beings provided the template for machine behaviour in one way or another.

228
00:23:31,310 --> 00:23:40,670
And what's interesting looking back, is that ultimately this approach of building machines in these different images of human beings did not succeed.

229
00:23:40,670 --> 00:23:45,350
And despite the initial burst of enthusiasm and optimism and excitement,

230
00:23:45,350 --> 00:23:51,500
really no progress was made in artificial intelligence by the really sort of serious noteworthy progress.

231
00:23:51,500 --> 00:23:55,940
And as progress faltered, researchers found themselves at a dead-end reset.

232
00:23:55,940 --> 00:24:01,190
Funding dried up. Progress came to an end. Interest in the field fell away.

233
00:24:01,190 --> 00:24:07,790
A period known as the A.I. winter began when really a not not a lot happened in the field at all.

234
00:24:07,790 --> 00:24:12,710
The great turning point, and it's a familiar moment, came in 1997 when Garry Kasparov,

235
00:24:12,710 --> 00:24:19,420
who at the time was the world chess champion, was beaten by a system owned by IBM called Deep Blue.

236
00:24:19,420 --> 00:24:24,670
What's so interesting about this, looking back on it, is that if you would ask my dad in the 1980s,

237
00:24:24,670 --> 00:24:30,070
do you think something like this will ever be possible? He would have said emphatically, no.

238
00:24:30,070 --> 00:24:35,680
And the reason he would have said no is very important. The reason he would have said no is because at the time they were on that first wave mindset,

239
00:24:35,680 --> 00:24:41,830
they thought the only way to build a system to outperform a human expert was to identify human experts.

240
00:24:41,830 --> 00:24:47,800
My dad's case with this man, Phillip Capper, sit down with him, get them to explain to you how it was.

241
00:24:47,800 --> 00:24:54,190
They had all that, every problem it was. You were trying to build a machine to build a machine to perform.

242
00:24:54,190 --> 00:25:00,980
And then you tried to capture the way that human beings were thinking and reasoning in a set of instructions for machine to follow.

243
00:25:00,980 --> 00:25:05,140
But here's the problem, and Garry Kasparov is a great example of this problem.

244
00:25:05,140 --> 00:25:09,340
If you sit down with Garry Kasparov and say, Garry, tell me how it is.

245
00:25:09,340 --> 00:25:14,800
You're so good at chess. Tell me what you're thinking about. Walk me through your reasoning processes.

246
00:25:14,800 --> 00:25:19,390
He might be able to give you a few clever opening moves or closing plays, but ultimately he'd struggle.

247
00:25:19,390 --> 00:25:24,490
He'd say things like It requires your instinct, intuition, gut reaction experience.

248
00:25:24,490 --> 00:25:28,660
I can't articulate how it is. I'm so good at chess.

249
00:25:28,660 --> 00:25:33,550
And for that reason, my dad and his colleagues thought something like this could never be automated of a human being.

250
00:25:33,550 --> 00:25:38,410
Can't articulate how it is. They perform a task. Where in us do we begin?

251
00:25:38,410 --> 00:25:42,840
They thought in writing a set of instructions for mission to follow.

252
00:25:42,840 --> 00:25:51,430
Of course, what they had a bank told in the 1980s was the exponential growth in processing power that would happen in the decades to come.

253
00:25:51,430 --> 00:25:56,530
So by the time the Garry Kasparov sat, and this is just to give you a sense of quite how powerful this is,

254
00:25:56,530 --> 00:26:01,660
that's a logarithmic scale there on the left hand side, which means each step up represents a ten fold increase.

255
00:26:01,660 --> 00:26:05,230
So one step, ten fold, two steps, one hundred fold, three steps, a thousand fold.

256
00:26:05,230 --> 00:26:13,000
So that is a really explosive growth in the computational power available for beginning in the 1950s,

257
00:26:13,000 --> 00:26:17,560
but really taking off in the middle of the second half of that century.

258
00:26:17,560 --> 00:26:21,670
So by the time Garry Kasparov sat down with deep blue in nineteen ninety seven,

259
00:26:21,670 --> 00:26:26,050
deep blue was calculating up to three hundred and thirty million moves a second.

260
00:26:26,050 --> 00:26:32,140
Garry Kasparov, at best, could maybe jog about one hundred and ten moves in his head on any one time he was blown

261
00:26:32,140 --> 00:26:36,850
out of the water by brute force purchasing power and lots of data storage capability.

262
00:26:36,850 --> 00:26:44,020
This system wasn't trying to replicate. His thinking process wasn't trying to mimic his reasoning.

263
00:26:44,020 --> 00:26:47,470
It was performing the task in a fundamentally different way.

264
00:26:47,470 --> 00:26:53,440
So the deep blue result was a practical victory, but it was also an ideological triumph as well.

265
00:26:53,440 --> 00:26:58,630
We can think of most A.I. researchers up until that point as being purists and closely

266
00:26:58,630 --> 00:27:04,090
observed human beings acting intelligently and trying to build machines like that.

267
00:27:04,090 --> 00:27:06,520
But that was not how Deep Blue was designed.

268
00:27:06,520 --> 00:27:14,860
Its creators didn't set out to copy the anatomy of human chess players, the reasoning they engaged in or the particular strategies they followed.

269
00:27:14,860 --> 00:27:16,390
They were pragmatists.

270
00:27:16,390 --> 00:27:24,100
They took a task that required intelligence work performed by a human being and built a machine to perform it in a fundamentally different way.

271
00:27:24,100 --> 00:27:27,190
And that's what brought that, and that's what brought the world of AI.

272
00:27:27,190 --> 00:27:35,890
I argue out of the air into what I call the pragmatist revolution, and a generation of systems is now being built in that spirit,

273
00:27:35,890 --> 00:27:46,330
crafted to function very differently from human beings, judged not by how they perform and task, but judged by how well they perform it.

274
00:27:46,330 --> 00:27:54,850
Advances in machine translation, for instance, have come not from developing a machine that mimics a talented translator, a human translator,

275
00:27:54,850 --> 00:28:01,510
but having computer scan millions of human translated pieces of text to figure out the different correspondences and patterns on the road.

276
00:28:01,510 --> 00:28:06,490
Likewise, machines have learnt to classify images not by mimicking human vision,

277
00:28:06,490 --> 00:28:14,800
but by reviewing millions of previously labelled pictures and hunting for similarities between those in the particular photo in question.

278
00:28:14,800 --> 00:28:18,100
I think this is also why, in the economics literature,

279
00:28:18,100 --> 00:28:26,440
we've seen a systematic underestimation of the capabilities of machines and what the tasks of driving a car,

280
00:28:26,440 --> 00:28:31,750
making a medical diagnosis and identifying a bird at a fleeting glimpse have in common.

281
00:28:31,750 --> 00:28:36,250
Well, these were all tasks that at one point leading economists thinking about the future of work,

282
00:28:36,250 --> 00:28:41,320
thinking about technology and its impact on work thought couldn't readily be automated.

283
00:28:41,320 --> 00:28:45,280
And yet today they increasingly can be. So what went wrong here?

284
00:28:45,280 --> 00:28:54,640
When again, I think these economists were purists believing that machines had to copy the way the human beings think and reason to outperform them?

285
00:28:54,640 --> 00:29:01,750
They were exactly the same mindset that my dad was, which was that if you want to build a system to outperform a human expert,

286
00:29:01,750 --> 00:29:06,220
you've got to sit down with them, get them to explain to you how it was. They do whatever task is.

287
00:29:06,220 --> 00:29:10,240
You're trying to build a machine to do and try and copy that human explanation.

288
00:29:10,240 --> 00:29:17,890
But if you ask a doctor, say, how is it that you make a medical diagnosis just like Garry Kasparov,

289
00:29:17,890 --> 00:29:22,210
they would probably struggle to explain exactly how it is that they do it.

290
00:29:22,210 --> 00:29:28,840
And this set of observations leads, I think, to one of the most important ideas in our in our work,

291
00:29:28,840 --> 00:29:33,980
which is the fallacy and the artificial intelligence fallacy.

292
00:29:33,980 --> 00:29:38,710
And it's this it's the mistaken assumption that the only way to develop systems to perform tasks

293
00:29:38,710 --> 00:29:43,570
at the level of human beings or higher is to copy the way that human beings perform that task.

294
00:29:43,570 --> 00:29:48,520
Now that was true in the first wave of artificial intelligence is simply no longer true,

295
00:29:48,520 --> 00:29:55,420
and I think it clouds so many of our judgements about what it is that systems of machines will be capable of doing in the future.

296
00:29:55,420 --> 00:29:58,630
So let me give you an example. Look, Daniel, you don't understand.

297
00:29:58,630 --> 00:30:05,860
Here's a job that requires judgement and judgement is the sort of thing that simply cannot be performed by machine.

298
00:30:05,860 --> 00:30:11,380
And again, in light of everything, I've just said the question Can a machine ever exercise judgement?

299
00:30:11,380 --> 00:30:15,880
Is the wrong question to be asking? In fact, there's two more important questions we should be asking.

300
00:30:15,880 --> 00:30:18,550
The first is to what problem is judgement the solution?

301
00:30:18,550 --> 00:30:24,320
Why do we go to our fellow human beings and say, Look, I need your judgement, how help me, I need you?

302
00:30:24,320 --> 00:30:28,420
Well, I think the answer to that question is because of uncertainty,

303
00:30:28,420 --> 00:30:32,380
when the facts are unclear, when the information is ambiguous, when we don't know what to do,

304
00:30:32,380 --> 00:30:40,090
we go to our fellow human beings and say, Look, I need your judgement based on your experience to help me make sense of this uncertainty.

305
00:30:40,090 --> 00:30:45,550
So really, the question we've got to be asking isn't in this case kind of machine ever exercised judgement,

306
00:30:45,550 --> 00:30:49,420
but it's kind of a machine to deal with uncertainty better than a human being.

307
00:30:49,420 --> 00:30:52,430
And the answer is in many. Of course they can. That just again,

308
00:30:52,430 --> 00:30:58,220
think of this system medical diagnostic system developed by a team of researchers at Stanford that

309
00:30:58,220 --> 00:31:03,230
can tell you whether or not to freckle is cancerous as accurately as leading dermatologists.

310
00:31:03,230 --> 00:31:07,100
How does it work? It's not trying to copy the judgement of a human doctor.

311
00:31:07,100 --> 00:31:11,060
It knows understands nothing about medicine at all.

312
00:31:11,060 --> 00:31:12,980
Instead, it's got a database of, I think,

313
00:31:12,980 --> 00:31:19,940
about one hundred and twenty nine thousand plus cases and is running essentially a pattern recognition algorithm through those cases.

314
00:31:19,940 --> 00:31:27,350
Hunting for similarities between them and the particular photo of a troubling freckle that you've given it is performing the task

315
00:31:27,350 --> 00:31:34,640
in an unhuman way based on the analysis of more possible cases than any human doctor could hope to review in their lifetime.

316
00:31:34,640 --> 00:31:42,680
It doesn't matter that a human doctor can't explain themselves how machines think love the question from a philosophical point of view.

317
00:31:42,680 --> 00:31:45,650
But again, in light of all this, in light of all I've said,

318
00:31:45,650 --> 00:31:56,990
I think it's not particularly helpful question to be asking when we think about the future of work to see why take a different system owned by IBM,

319
00:31:56,990 --> 00:32:01,520
a system called Watson. It's claim to fame, of course, that was that.

320
00:32:01,520 --> 00:32:09,260
It went on the US quiz show Jeopardy in 2011, and it beat the two human champions at Jeopardy.

321
00:32:09,260 --> 00:32:17,240
And again, it was a remarkable achievement. What I loved about this, though, looking back on it, was that the day after Watson won on Jeopardy,

322
00:32:17,240 --> 00:32:22,190
the Wall Street Journal ran a piece by the great philosopher John Self with the title.

323
00:32:22,190 --> 00:32:26,450
Watson doesn't know it won a jeopardy. Right?

324
00:32:26,450 --> 00:32:28,220
And it's brilliant. That is completely true.

325
00:32:28,220 --> 00:32:33,360
You know, Watson didn't let out a point of excitement, didn't call up his parents to say what a good job I had done.

326
00:32:33,360 --> 00:32:36,650
You know, didn't want to go down to the proverbial pub to have a drink with his friends.

327
00:32:36,650 --> 00:32:42,140
The system wasn't trying to copy the way that these human contestants thought or reasoned, but it didn't matter.

328
00:32:42,140 --> 00:32:47,120
It still outperformed them. And that's what I think is what the second wave of artificial intelligence is about.

329
00:32:47,120 --> 00:32:52,940
And that's where we find ourselves today. Systems and machines, which are using lots of processing power,

330
00:32:52,940 --> 00:32:58,340
lots of data storage capability and advances in algorithms designed to perform tasks that might

331
00:32:58,340 --> 00:33:03,920
require faculties like judgement and creativity and empathy when performed by human beings,

332
00:33:03,920 --> 00:33:07,860
but are able now to perform those tasks in fundamentally different ways.

333
00:33:07,860 --> 00:33:13,190
And so a whole realm of tasks that we thought were once out of reach of automation.

334
00:33:13,190 --> 00:33:19,220
More and more of those tasks appear to be within reach. So what does this then mean for thinking about what?

335
00:33:19,220 --> 00:33:23,680
What are these changes in the world of technology mean for thinking about work?

336
00:33:23,680 --> 00:33:33,550
Greek mythology tells the story of a man called tenseness who chops up his son and serves him as a meal to the gods.

337
00:33:33,550 --> 00:33:40,660
This, given his dinner guests omniscience, turns out to be a very bad decision, and as a punishment,

338
00:33:40,660 --> 00:33:49,510
he's made to stand for eternity in a pool of water up to his chin, surrounded by trees bursting with fruit.

339
00:33:49,510 --> 00:33:53,830
But each time he leans down to take a sip, the water recedes away from his lips.

340
00:33:53,830 --> 00:33:59,810
And every time he reaches out to take some fruit from the tree, the branches swing away from his grasp.

341
00:33:59,810 --> 00:34:02,740
And this, I think, captures.

342
00:34:02,740 --> 00:34:09,940
This story of 10 summers, which gives us the word tantalise captures, I think the spirit of the first kind of technological unemployment,

343
00:34:09,940 --> 00:34:15,460
which we can think of as being a sort of frictional technological unemployment here.

344
00:34:15,460 --> 00:34:22,990
There's still work to be done. The problem is that not all workers are able to reach out and take it up.

345
00:34:22,990 --> 00:34:28,890
So friction a of technological unemployment doesn't mean that there's going to be fewer jobs for human beings to do.

346
00:34:28,890 --> 00:34:37,030
It's important to remember that if we think again in terms of those two forces, I think for the next decade or so, at least in almost all economies,

347
00:34:37,030 --> 00:34:42,280
that harmful substituting force that displaces workers will be overwhelmed by

348
00:34:42,280 --> 00:34:46,900
the helpful complementing force that raises the demand for that work elsewhere.

349
00:34:46,900 --> 00:34:53,140
I think that's going to continue. That historical story will continue. But for three reasons, and I set these out in the book,

350
00:34:53,140 --> 00:35:01,570
I think that this in-demand work is increasingly going to be out of the reach of more and more people if we want to take it up.

351
00:35:01,570 --> 00:35:05,230
What are these three reasons? The first is perhaps the most familiar.

352
00:35:05,230 --> 00:35:09,730
It's the skills mismatch where displaced workers do not necessarily have the skills

353
00:35:09,730 --> 00:35:14,350
and capabilities to do the new work that's created by technological progress.

354
00:35:14,350 --> 00:35:18,310
And this is probably the most familiar reason why people might not be able to

355
00:35:18,310 --> 00:35:23,680
take up the work that's available and how we can perhaps talk about it later on.

356
00:35:23,680 --> 00:35:31,990
The second reason is a geographical one place mismatch where displaced workers do not live in the same place as the new work that is created.

357
00:35:31,990 --> 00:35:35,230
I mean, it's interesting if we think back to the early days of the internet,

358
00:35:35,230 --> 00:35:39,280
there was a moment where it seemed like these sorts of worries about location might no

359
00:35:39,280 --> 00:35:45,070
longer matter that people spoke about the death of distance and how the world is flat.

360
00:35:45,070 --> 00:35:52,160
But it turns out that really, in looking for work today, the place where you live matters more than ever.

361
00:35:52,160 --> 00:35:57,860
I think the third reason for frictional technological unemployment is perhaps the least familiar.

362
00:35:57,860 --> 00:36:06,380
And it's what I call identity mismatch, and this is where displaced workers have an identity rooted in a particular sort

363
00:36:06,380 --> 00:36:11,840
of work and are willing to stay unemployed in order to protect the identity.

364
00:36:11,840 --> 00:36:18,410
So think of adult men in the United States, for instance, displaced from manufacturing roles by new technologies.

365
00:36:18,410 --> 00:36:27,080
Some say that they prefer not to work at all than to take up that it's an unfortunate term take up so-called pink collar work.

366
00:36:27,080 --> 00:36:33,950
And it's a term that's meant to capture the fact that many of the roles currently out of reach of machines are disproportionately held by women.

367
00:36:33,950 --> 00:36:40,940
So ninety seven point seven percent of pre-school and kindergarten teachers in the United States are women.

368
00:36:40,940 --> 00:36:47,420
Ninety two point two percent of nurses are women. Eighty two point five percent of social workers are women.

369
00:36:47,420 --> 00:36:54,650
So I think together those three mismatches pay to tell a story where there might be work for people to do.

370
00:36:54,650 --> 00:37:00,520
But I think there are important worries. Two reasons to think that there might not be able to necessarily take it up.

371
00:37:00,520 --> 00:37:06,620
And I think most economists tend to be comfortable with this idea of friction or technological unemployment.

372
00:37:06,620 --> 00:37:13,100
I think many of them can picture a world where there's lots of work to be done because of these mismatches.

373
00:37:13,100 --> 00:37:20,750
Not everyone can do it. But as we move through the 21st century, I think we might see the emergence of a second type of technological unemployment,

374
00:37:20,750 --> 00:37:25,580
one that's not set, one where there simply isn't enough well-paid work to be done.

375
00:37:25,580 --> 00:37:33,560
Full stop. And I call this structural technological unemployment, and this is, I think, is a less comfortable idea.

376
00:37:33,560 --> 00:37:42,950
So can this be right? Can you? What about the fact that after 300 years of radical technological change, there's still enough work for people to do?

377
00:37:42,950 --> 00:37:49,860
Does that not tell us there's always going to be enough work, enough demand for the work of human beings?

378
00:37:49,860 --> 00:37:53,870
What I try and argue in the book is that, no, that's not the case,

379
00:37:53,870 --> 00:38:00,590
and the fundamental reason why is that process of tasking encroachment that I described before.

380
00:38:00,590 --> 00:38:07,160
So we think again, in terms of those two forces, you know, there can be little doubt is that as task encroachment continues,

381
00:38:07,160 --> 00:38:15,050
that harmful substituting force is going to grow stronger. The workers are going to be displaced from a wider range of tasks than ever before.

382
00:38:15,050 --> 00:38:21,350
So the key question is why can't we simply rely on that helpful complementing force to overcome that effect,

383
00:38:21,350 --> 00:38:30,020
as it has done since modern economic growth began? And the answer, I think, is that task encroachment also has a second pernicious effect,

384
00:38:30,020 --> 00:38:36,770
which is that over time, I worry that it might wear down that helpful complementing force as well.

385
00:38:36,770 --> 00:38:43,340
So let's just think about that complementing force a little more. Take the productivity effect that I mentioned originally in the future.

386
00:38:43,340 --> 00:38:48,980
New technologies are no doubt going to make some people more productive assets and tasks.

387
00:38:48,980 --> 00:38:51,800
That's almost certainly going to happen.

388
00:38:51,800 --> 00:38:58,430
But this is only going to continue to help workers if they remain better placed to do those tasks than machines.

389
00:38:58,430 --> 00:39:03,550
And as task encroachment continues and continues, that becomes less and less likely for more and more tasks.

390
00:39:03,550 --> 00:39:09,320
So you think about satnav systems again, just to see what I'm getting at today, as I said before,

391
00:39:09,320 --> 00:39:15,260
they make it easier for taxi drivers to navigate unfamiliar roads, making them better at the wheel.

392
00:39:15,260 --> 00:39:19,340
But that's only going to be true so long as human beings are better placed.

393
00:39:19,340 --> 00:39:27,290
The machines to steer a vehicle from A to B and in the coming years of excitement about driverless cars is right.

394
00:39:27,290 --> 00:39:33,530
That's no longer going to be the case. Your software is likely to drive cars more efficiently and safely than us.

395
00:39:33,530 --> 00:39:39,920
And at that point, it's no longer going to matter how good people are at driving with or without signs.

396
00:39:39,920 --> 00:39:45,410
The machines will simply do it instead. I'll think about the bigger pie effect in the future.

397
00:39:45,410 --> 00:39:53,150
Again, economic plays will no doubt continue to grow. Incomes will be larger than ever before, and demand for goods is going to soar.

398
00:39:53,150 --> 00:39:58,580
It can be pretty certain about that. Yeah, I don't think we can necessarily rely.

399
00:39:58,580 --> 00:40:05,270
I don't think we can necessarily rely on this to bolster the demand for the work of human beings, as it has in the past.

400
00:40:05,270 --> 00:40:13,820
Why? Because just as with that productivity effect, the bigger pie effect is only going to help if people rather than machines remain

401
00:40:13,820 --> 00:40:18,200
better placed to perform whatever tasks have to be done to produce those goods.

402
00:40:18,200 --> 00:40:22,010
And again, it's task encroachment continues that becomes less and less likely,

403
00:40:22,010 --> 00:40:27,110
and we can already see something like this, in particular corners of economic life.

404
00:40:27,110 --> 00:40:30,860
So think about UK agriculture UK agricultural sector.

405
00:40:30,860 --> 00:40:39,770
This part of the British economic pie has grown dramatically over the last century and a half, but it's simply not created more work for people to do.

406
00:40:39,770 --> 00:40:45,420
British agriculture now produces about five times what it did back in 1860.

407
00:40:45,420 --> 00:40:53,060
Yeah, only requires a tenth of the workers to do it. Think about UK manufacturing a similar story since 1948.

408
00:40:53,060 --> 00:40:57,470
Sector now produces about one hundred and fifty percent more than it did back then.

409
00:40:57,470 --> 00:41:00,530
It requires 60 percent fewer workers to do it.

410
00:41:00,530 --> 00:41:09,440
A growing economic pie doesn't necessarily mean in these particular corners of the economy growing demand for the work of human beings as well.

411
00:41:09,440 --> 00:41:11,660
And finally, think about that changing pie effect.

412
00:41:11,660 --> 00:41:18,590
You know, again, the economic pie may change, but in exactly the same way as this process of task encroachment continues,

413
00:41:18,590 --> 00:41:25,430
it becomes more and more likely that machines, rather than human beings, will be better placed to do whatever new tasks have to be done.

414
00:41:25,430 --> 00:41:31,910
And again, you know, if you look at you parts of economic life, you might worry that something like this is already unfolding.

415
00:41:31,910 --> 00:41:41,930
So in nineteen sixty four, the most valuable company in the United States was AT&T, with seven hundred and seven hundred and sixty thousand employees.

416
00:41:41,930 --> 00:41:47,510
Roll forward those of twenty eighteen and as Apple with only one hundred and thirty two thousand employees.

417
00:41:47,510 --> 00:41:52,400
Twenty nineteen. It was Microsoft with only one hundred and thirty one thousand.

418
00:41:52,400 --> 00:42:00,320
The economy has transformed when we look at some of the leading companies, but it's those companies require fewer workers.

419
00:42:00,320 --> 00:42:05,090
I mean, the more general point is this, which is that in the year 2000,

420
00:42:05,090 --> 00:42:12,200
you industries that were created in the 21st century accounted for just half a percent of U.S. employment.

421
00:42:12,200 --> 00:42:21,600
Clearly, the U.S. economy has transformed, but that hasn't. Assertive use necessarily created lots and lots more work for people to do.

422
00:42:21,600 --> 00:42:28,400
So this, I think, gives you a flavour of the argument that I'm making in the book about how we might find ourselves in a world with less work.

423
00:42:28,400 --> 00:42:34,880
As time goes on, machines continue to become more and more capable, taking on tasks that once fell to human beings.

424
00:42:34,880 --> 00:42:38,810
The harmful substituting force displaces workers in the familiar way.

425
00:42:38,810 --> 00:42:45,800
And for a time, I think the helpful complementing force is going to continue to raise the demand for those displaced workers elsewhere.

426
00:42:45,800 --> 00:42:52,100
And I challenge in that world is going to be frictional technological unemployment.

427
00:42:52,100 --> 00:42:56,750
But a toxic encroachment continues and more and more tasks for to machines.

428
00:42:56,750 --> 00:43:00,140
I worry that the hopeful complementing force is going to be can.

429
00:43:00,140 --> 00:43:06,500
As well, the human beings will find themselves confined to an ever shrinking set of tasks and activities.

430
00:43:06,500 --> 00:43:13,070
And there's no reason, no economic law that tells us there must necessarily be enough demand for those

431
00:43:13,070 --> 00:43:18,080
residual activities to keep everyone who wants it and well-paid employment.

432
00:43:18,080 --> 00:43:22,610
And then our challenge becomes structural technological unemployment.

433
00:43:22,610 --> 00:43:25,370
And so the world of work comes to an end, not with some Big Bang,

434
00:43:25,370 --> 00:43:31,520
not with robots taking everyone's jobs in a flash in an instant, but with a sort of gradual withering,

435
00:43:31,520 --> 00:43:39,770
a withering in the demand for the work of human beings, as that substituting force gradually overruns the helpful complimenting force

436
00:43:39,770 --> 00:43:45,710
and the balance between the two no longer tips in favour of human beings.

437
00:43:45,710 --> 00:43:52,610
So what problems does this lead to? I just want to sketch out what I think are the main challenges that we're going to face as a result of this,

438
00:43:52,610 --> 00:43:57,410
and we can perhaps discuss what we do about these in more detail in a moment.

439
00:43:57,410 --> 00:44:01,120
The first is the economic challenge, the challenge of inequality.

440
00:44:01,120 --> 00:44:07,100
You know, today the labour market is the main way that we share our economic prosperity and society,

441
00:44:07,100 --> 00:44:11,240
that most people's jobs are their only source of income.

442
00:44:11,240 --> 00:44:18,290
How do we share out material prosperity in a society where our traditional way of doing so came?

443
00:44:18,290 --> 00:44:22,850
People for the work that they do is far less effective than in the past.

444
00:44:22,850 --> 00:44:26,150
How do we do it? What's going to take the labour market place?

445
00:44:26,150 --> 00:44:31,730
I see these worries about a world with less work as being very closely related to worries about inequality.

446
00:44:31,730 --> 00:44:34,910
Today, the two things are very, very closely related.

447
00:44:34,910 --> 00:44:40,490
So I think the economic challenge in the 21st century is increasingly going to be a distributional one.

448
00:44:40,490 --> 00:44:50,420
But I don't think the challenges are just about economics. A big second challenge is what we do about the growing power of large technology

449
00:44:50,420 --> 00:44:56,120
companies who are increasingly responsible for developing these technologies.

450
00:44:56,120 --> 00:45:03,110
And I think what's interesting is that the 20th century, our main worry about large corporations was their economic power.

451
00:45:03,110 --> 00:45:06,500
You know, we focussed on things like profit and crisis.

452
00:45:06,500 --> 00:45:13,130
And we used competition policy and economists told us whether or not the economic power was excessive or not.

453
00:45:13,130 --> 00:45:19,460
I think in the 21st century, our worry is going to be far less about the economic power of large technology companies

454
00:45:19,460 --> 00:45:23,480
and far more about their political power is going to be far more about issues,

455
00:45:23,480 --> 00:45:32,120
liberty and democracy and social justice, things that have often very little to do with prices and profits.

456
00:45:32,120 --> 00:45:34,220
And it might sound like I'm shooting myself in the foot,

457
00:45:34,220 --> 00:45:40,820
but things that economists who've traditionally dominated these discussions aren't necessarily best placed to talk about.

458
00:45:40,820 --> 00:45:48,710
And so I think that is the challenge for the 21st century when we think about large companies, what we do about that growing political power.

459
00:45:48,710 --> 00:45:53,780
The final challenge is the challenge of meaning. What do we do?

460
00:45:53,780 --> 00:46:00,440
How do we provide meaning and purpose in people's lives when work?

461
00:46:00,440 --> 00:46:04,700
For many people, a traditional source of meaning and purpose is no longer available?

462
00:46:04,700 --> 00:46:12,920
What do we do? And that, I think, is again a challenge that doesn't have much to do with economics,

463
00:46:12,920 --> 00:46:19,650
but it really is fundamental to the nature of the problems that we're going to face if we move into a world with less work.

464
00:46:19,650 --> 00:46:21,540
So those are the challenges I think we're going to face.

465
00:46:21,540 --> 00:46:28,080
Let me just in the final minutes explain though, why in spite of all of this, I do remain optimistic.

466
00:46:28,080 --> 00:46:32,340
And the reason I think is simple, which is that in decades to come,

467
00:46:32,340 --> 00:46:39,690
technological progress is going to solve or is likely to solve the economic problem that's dominating humanity until now.

468
00:46:39,690 --> 00:46:42,390
And if we think again, if the economy isn't pie,

469
00:46:42,390 --> 00:46:49,740
the traditional challenge has been how do we make that pie large enough everyone to live on at the turn of the first century A.D.,

470
00:46:49,740 --> 00:46:54,420
if you had taking the global economic pie and divided up into equal slices for everyone in the world,

471
00:46:54,420 --> 00:46:57,060
most people would have got a few hundred dollars.

472
00:46:57,060 --> 00:47:04,260
Almost everyone lived on or around the poverty line roll forward a thousand years, and roughly the same was true.

473
00:47:04,260 --> 00:47:08,310
But as we saw on those opening slides over the last few hundred years,

474
00:47:08,310 --> 00:47:14,190
economic growth has soared and this growth was of course driven in large part by technological progress.

475
00:47:14,190 --> 00:47:17,790
Economic policies around the world have got far bigger today.

476
00:47:17,790 --> 00:47:24,480
If we take the global economic pie and divided up into equal slices for everyone in the world, everyone gets about $11000.

477
00:47:24,480 --> 00:47:35,270
Thirty years, it'll be double that in 60 years. Double that again. It by and large, we come very close to solving the traditional economic problem,

478
00:47:35,270 --> 00:47:41,780
the struggle for subsistence, as Keynes called that has dominated humankind until now.

479
00:47:41,780 --> 00:47:49,280
And technological unemployment in a strange way, I see as a symptom of success a symptom of that success.

480
00:47:49,280 --> 00:47:53,090
The 21st century technological progress is going to solve one problem,

481
00:47:53,090 --> 00:47:59,210
how do we make the pie large enough for everyone to live on that is going to replace it with those three others?

482
00:47:59,210 --> 00:48:05,180
These problems of inequality, problems of power and problems of meaning and purpose.

483
00:48:05,180 --> 00:48:12,350
And clearly, I think there's going to be huge disagreement about how we meet these challenges about how we should share our prosperity and society,

484
00:48:12,350 --> 00:48:19,280
about how we constrain the political power of large technology companies and how we provide meaning in a world with less work.

485
00:48:19,280 --> 00:48:22,820
But I think these are and this is the argument I make in the book.

486
00:48:22,820 --> 00:48:26,780
These are in the final analysis, far more attractive problems, I believe,

487
00:48:26,780 --> 00:48:32,450
to have to grapple with than the ones the hope with and the wonderful things that our ancestors for centuries,

488
00:48:32,450 --> 00:48:38,690
which was how to make that point big enough in the first place. So I will finish there.

489
00:48:38,690 --> 00:48:44,940
Thank you very much for your attention, and I look forward to hearing some reflections and taking some questions as well.

490
00:48:44,940 --> 00:48:57,670
Thank you. Thank you very much.

491
00:48:57,670 --> 00:49:07,390
You've heard wonderfully lucid summary of your book and clearly just touching on some of the key things, but hopefully enticing us all to read it.

492
00:49:07,390 --> 00:49:14,620
I certainly have and commend to all of you. I think one person who's guaranteed to have a job in the future is here.

493
00:49:14,620 --> 00:49:19,570
I don't see a machine doing anything as formidable as what you've just done.

494
00:49:19,570 --> 00:49:27,820
And incidentally, if interesting in Garry Kasparov, view on what that was all about with deep blue challenging him.

495
00:49:27,820 --> 00:49:32,800
You know, he's been a fellow here, visiting fellow and somewhere on a video archive.

496
00:49:32,800 --> 00:49:36,900
You'll find him talking about that experience and what he thinks about the future.

497
00:49:36,900 --> 00:49:46,340
And so we have about time for very few questions, but who'd like to go first?

498
00:49:46,340 --> 00:49:52,430
Maybe we should collectively act. Thank you very much.

499
00:49:52,430 --> 00:49:57,480
I think he touched on what I'm about to ask in the last few minutes about the distributional issues.

500
00:49:57,480 --> 00:50:03,380
Yes. But earlier you said demand for goods is going to soar.

501
00:50:03,380 --> 00:50:11,550
Yes. But I'm wondering who's going to demand those goods from a kind of consumption focus rather than a production focus?

502
00:50:11,550 --> 00:50:19,440
We're just going to collect a couple of questions post them here. She's making down.

503
00:50:19,440 --> 00:50:28,110
And John Hofmeyr from the Centre for Mutual and Cohn Business here at Oxford, as you talk about inequality.

504
00:50:28,110 --> 00:50:33,090
One of the solutions people suggest is a basic income.

505
00:50:33,090 --> 00:50:42,720
My suggestion is that that won't really help overcome inequality, and it's really at the base that wealth is produced by ownership of shares.

506
00:50:42,720 --> 00:50:55,890
I wonder if you could just comment on whether you've thought about that and the personal view of the two people on that story where they.

507
00:50:55,890 --> 00:51:03,820
Hey, I'm just curious, what do you think the implications of a world without work will be for challenges such as climate change?

508
00:51:03,820 --> 00:51:08,790
I'm, uh, my name's Jasmine, I'm from Stanford University.

509
00:51:08,790 --> 00:51:14,760
So I guess I'm wondering. It seems like work provides a set of intrinsic benefits to the job holder, such as meaning,

510
00:51:14,760 --> 00:51:20,520
skills, et cetera, as well as a set of extrinsic benefits like training people to follow rules,

511
00:51:20,520 --> 00:51:21,840
keeping people off the streets,

512
00:51:21,840 --> 00:51:28,530
corporate provided health care and a host of other sort of public goods that having our current system of work provides.

513
00:51:28,530 --> 00:51:32,790
So I'm so it seems that widespread unemployment would threaten the centralised political

514
00:51:32,790 --> 00:51:37,530
power and the ability to keep regime stable because of those extrinsic benefits.

515
00:51:37,530 --> 00:51:42,140
I'm curious how you think that will play out or how governments might be able to manage it.

516
00:51:42,140 --> 00:51:48,680
Hmm. Yeah, so let me try. Let me try and respond to some of those questions at the same time.

517
00:51:48,680 --> 00:51:56,520
I mean, the first two, just again, I see the economic challenge that we face as a distributional one.

518
00:51:56,520 --> 00:52:02,020
Yeah, the labour market has traditionally been our way of sharing our prosperity and society.

519
00:52:02,020 --> 00:52:08,740
If we are approaching a world with less work than that labour, then that mechanism that we've relied upon will be less effective than in the past.

520
00:52:08,740 --> 00:52:11,020
And so the question is, well, what will replace it?

521
00:52:11,020 --> 00:52:18,430
The argument that I make in the book is that I think we're going to see a growing role of the state in solving that problem.

522
00:52:18,430 --> 00:52:21,460
And I call it the big state, not the big state of the 20th century.

523
00:52:21,460 --> 00:52:28,750
Teams of smart people sitting with blueprints and central government trying to command and control entire economies and how they operate,

524
00:52:28,750 --> 00:52:36,130
not in production, but in distribution. And I say that not coming from the left or right, just looking at the technical features,

525
00:52:36,130 --> 00:52:42,700
the problem and just saying if the labour market is less effective at sharing prosperity, then we'll need the state to do it in its place.

526
00:52:42,700 --> 00:52:50,740
And universal basic income is one thing that's proposed in the book Occupy critical of Universal Basic Income, and I propose quite different ideas.

527
00:52:50,740 --> 00:52:54,540
But I think the general.

528
00:52:54,540 --> 00:53:03,750
I think this is I think it's important to have debates and conversations about UBI because they are engaging with the based on the right diagnosis,

529
00:53:03,750 --> 00:53:04,470
which is,

530
00:53:04,470 --> 00:53:16,370
well, how can we maybe share our income in society if we can't rely on the labour market to do the meaning and purpose component of this problem is.

531
00:53:16,370 --> 00:53:18,860
And I spent a lot of time in the book thinking about, I think it's very important.

532
00:53:18,860 --> 00:53:26,660
I mean, just in response to your particular observation, one of the things I try and do is really explore this relationship between work and meaning.

533
00:53:26,660 --> 00:53:32,810
I think we often assume that work is an important source of meaning and purpose for everyone when actually,

534
00:53:32,810 --> 00:53:38,180
if you look at the data, lots of people don't get a sense of identity and purpose from their work.

535
00:53:38,180 --> 00:53:45,470
Lots of people don't think that their work makes a meaningful contribution to the world that we live in.

536
00:53:45,470 --> 00:53:51,560
And if you look back in history, you can see very different relationships between work and meaning.

537
00:53:51,560 --> 00:54:03,560
So in the ancient Egyptian city of Thebes, you were banned from seeking political office if you had engaged in trade in work in the previous decade.

538
00:54:03,560 --> 00:54:12,710
In Sparta, citizens were banned by law from engaging in productive work.

539
00:54:12,710 --> 00:54:16,850
You know, Aristotle and Plato both thought that work was a sort of prohibitively grubby affair.

540
00:54:16,850 --> 00:54:22,670
That meaning and purpose could only really come from certain types of leisure activities, and we could talk more about that.

541
00:54:22,670 --> 00:54:23,600
But that's the sort of thing.

542
00:54:23,600 --> 00:54:32,880
So I think the first thing we've got to do is begin with this assumption that many of us have that work is a source of meaning and purpose.

543
00:54:32,880 --> 00:54:36,690
For everyone to be whether or not there might be alternative sources of meaning and

544
00:54:36,690 --> 00:54:42,390
purpose and and if your answer to that first question is it's not for everyone to be,

545
00:54:42,390 --> 00:54:50,310
there might be other sources of it, then it opens up a whole fascinating, I think, new set of policy possibilities, not about the future of what,

546
00:54:50,310 --> 00:54:54,600
but about the future of leisure, not about labour market policies, but about leisure policies.

547
00:54:54,600 --> 00:54:59,620
And how do we as a society shape how we spend our time when we're not at work?

548
00:54:59,620 --> 00:55:05,760
It sounds quite radical, but we already do it in many ways today. Entrancing entrance to museums is free.

549
00:55:05,760 --> 00:55:15,600
We try and stop the fine artworks from leaving the country. We encourage young children to learn to psycho and sweat and swim.

550
00:55:15,600 --> 00:55:21,690
The pension system, in a way you could look at it and think, Well, you know, that's a form of leisure policy is essentially the state saying,

551
00:55:21,690 --> 00:55:26,490
Where are you really going to really support leisure in the twilight of life?

552
00:55:26,490 --> 00:55:31,200
That's when the only moment we're going to really provide financial support for leisure.

553
00:55:31,200 --> 00:55:38,070
So I think as a whole, a fascinating set of policy possibilities that open up when we start looking more carefully at this meaning and purpose.

554
00:55:38,070 --> 00:55:41,610
Part of the problem, and that's what I'm trying to do in the book. Climate change is very interesting.

555
00:55:41,610 --> 00:55:53,220
It's not something I look at in in the explicitly in the book, but I see it as being a challenge on the same sort of scale.

556
00:55:53,220 --> 00:55:57,600
I mean, you know, if we had had one of the reasons I've written a book is that because I think

557
00:55:57,600 --> 00:56:02,400
that if we had had the right conversations about climate change 30 years ago,

558
00:56:02,400 --> 00:56:06,630
we wouldn't be necessarily in the problem facing the problems that we face today.

559
00:56:06,630 --> 00:56:12,330
And I think I think of this problem in a similar way that these are conversations about inequality and what we do about it,

560
00:56:12,330 --> 00:56:19,730
that we need to engage with today to avoid finding ourselves in very difficult problems in a few decades time.

561
00:56:19,730 --> 00:56:25,800
We've got time for one more question. If I would, actually I was just going to a Mike.

562
00:56:25,800 --> 00:56:36,520
Yes. Mike, Mike, Mike. By I way, this is being webcast with the headset on those questions if you don't want to.

563
00:56:36,520 --> 00:56:40,770
And it was a bit about your it was following on from your point about work as a way of

564
00:56:40,770 --> 00:56:46,950
distributing income because we need income to get access to resources and services.

565
00:56:46,950 --> 00:56:57,580
But actually, air is going to in many ways will it will it will lower the cost, the marginal cost of producing many services that we require now.

566
00:56:57,580 --> 00:57:01,620
We won't be relying on expensive expertise.

567
00:57:01,620 --> 00:57:03,580
That will just be what you know,

568
00:57:03,580 --> 00:57:12,090
the access to medical diagnosis will be dependent on what those companies who hold power in the knowledge choose to charge us.

569
00:57:12,090 --> 00:57:15,520
But the actual cost of it will be the marginal cost will be very low.

570
00:57:15,520 --> 00:57:21,690
So actually, we could provide for people very, sort of cheaply in a way.

571
00:57:21,690 --> 00:57:28,290
I mean, it could almost be a quid pro quo for those companies that they provide housing, they provide transport,

572
00:57:28,290 --> 00:57:35,070
they provide energy for people because marginal cost of doing that should be much lower.

573
00:57:35,070 --> 00:57:38,490
And this was some of what you're observing.

574
00:57:38,490 --> 00:57:43,680
There was the argument that I made in the book that I co-authored with my dad the future of the professions,

575
00:57:43,680 --> 00:57:48,990
arguing that the promise of a lot of these technologies is a liberation of the sorts

576
00:57:48,990 --> 00:57:53,790
of expertise that traditionally has only been available to a very privileged, a very lucky few people.

577
00:57:53,790 --> 00:58:01,520
Access to a good health care system, access to knowledge of their legal entitlements, access to knowledge of how to manage their financial affairs.

578
00:58:01,520 --> 00:58:05,850
Suppose the observation we make in that book, though, is that this doesn't happen automatically and that you know,

579
00:58:05,850 --> 00:58:13,530
there's a there's a fight to be had to make sure that this expertise is widely available and in theory, correct.

580
00:58:13,530 --> 00:58:21,240
Well, it's always good to end the conversation like this with many more questions than answers than it will be around signing books,

581
00:58:21,240 --> 00:58:27,030
but also hopefully have time to respond to the many of you that haven't had a chance to answer your questions.

582
00:58:27,030 --> 00:58:35,760
And as I said, this is a wonderfully provocative and I think significant book and part of a much bigger debate,

583
00:58:35,760 --> 00:58:43,110
which the Oxford Martin School is certainly engaged in on the future of work and economies.

584
00:58:43,110 --> 00:58:47,490
It's, of course, engaged in many other things on the 28th of February.

585
00:58:47,490 --> 00:58:57,240
We have a whole series of discussions panels chaired after a keynote address by Andy Haldane,

586
00:58:57,240 --> 00:59:03,570
who's the chief economist at the Bank of England, and every Thursday in term time at five o'clock.

587
00:59:03,570 --> 00:59:07,560
There's a seminar series as well, so I encourage you to look at the websites.

588
00:59:07,560 --> 00:59:15,300
And if you want to ping tonight's, talk to your friends and others that haven't had a chance to be here or to look at it again.

589
00:59:15,300 --> 00:59:42,165
That will be up on our website and thanks to you all for coming. And please do join us for a drink and in thanking Daniel.