1
00:00:00,090 --> 00:00:02,430
The following content is
provided under a Creative

2
00:00:02,430 --> 00:00:03,820
Commons license.

3
00:00:03,820 --> 00:00:06,030
Your support will help
MIT OpenCourseWare

4
00:00:06,030 --> 00:00:10,120
continue to offer high-quality
educational resources for free.

5
00:00:10,120 --> 00:00:12,660
To make a donation or to
view additional materials

6
00:00:12,660 --> 00:00:16,620
from hundreds of MIT courses,
visit MIT OpenCourseWare

7
00:00:16,620 --> 00:00:17,992
at ocw.mit.edu.

8
00:00:22,360 --> 00:00:26,910
WILLIAM BONVILLIAN: This video
that we saw by David Mindell

9
00:00:26,910 --> 00:00:30,270
is about his new
book, which is called

10
00:00:30,270 --> 00:00:33,390
Our Robots Ourselves, Robots
and the Myth of Autonomy.

11
00:00:33,390 --> 00:00:35,250
And he means exactly that.

12
00:00:35,250 --> 00:00:39,540
Autonomy really is a
myth and will be so.

13
00:00:39,540 --> 00:00:42,540
And David has a unique
perspective on this question,

14
00:00:42,540 --> 00:00:45,728
because this is his life.

15
00:00:45,728 --> 00:00:47,520
In other words, he has
been deeply involved

16
00:00:47,520 --> 00:00:50,730
with robotics at just
cutting-edge levels

17
00:00:50,730 --> 00:00:53,800
for the last 30-plus years.

18
00:00:53,800 --> 00:00:57,360
So he was deeply involved in
a lot of the early undersea

19
00:00:57,360 --> 00:01:01,590
exploration and working
with Bob Ballard and Alvin

20
00:01:01,590 --> 00:01:03,900
and the pods that
came off Alvin.

21
00:01:03,900 --> 00:01:08,070
And then he did a
lot of work with NASA

22
00:01:08,070 --> 00:01:11,820
and was involved in
the robotics on Mars

23
00:01:11,820 --> 00:01:16,980
and the planetary
explorations using robotics.

24
00:01:16,980 --> 00:01:23,970
And then he spent a lot of
time with the US military

25
00:01:23,970 --> 00:01:25,740
that had been doing
drone technology

26
00:01:25,740 --> 00:01:27,630
and has a real hands-on feel.

27
00:01:27,630 --> 00:01:30,180
He's also a pilot and
flies extensively,

28
00:01:30,180 --> 00:01:33,510
both helicopters and aircraft.

29
00:01:33,510 --> 00:01:36,040
He's been involved in a
helicopter-drone company,

30
00:01:36,040 --> 00:01:39,810
so he knows exactly what
those are and how they work.

31
00:01:39,810 --> 00:01:44,130
And in addition, he's
just taken leave from MIT,

32
00:01:44,130 --> 00:01:46,830
and he set up his own robotics
company, including with people

33
00:01:46,830 --> 00:01:50,370
from Lincoln Labs, which has
the greatest radar expertise.

34
00:01:50,370 --> 00:01:51,960
And the issue there
is, how do you

35
00:01:51,960 --> 00:01:54,930
create robots that can work
in really intimate proximity

36
00:01:54,930 --> 00:01:56,760
with people, right?

37
00:01:56,760 --> 00:01:59,550
Not kind of arm's length
distance, much less

38
00:01:59,550 --> 00:02:04,113
locked in on a floor, but
really in close proximity.

39
00:02:04,113 --> 00:02:06,030
So they're developing
interesting technologies

40
00:02:06,030 --> 00:02:06,780
to enable that.

41
00:02:06,780 --> 00:02:11,790
So he's got a whole personal
sense of what this robotics

42
00:02:11,790 --> 00:02:13,350
movement is all about.

43
00:02:13,350 --> 00:02:19,350
And he kind of takes us back
from the literature that

44
00:02:19,350 --> 00:02:22,620
has portrayed this kind of
ongoing nightmare of job

45
00:02:22,620 --> 00:02:24,750
displacement from robotics,
and I think, frankly,

46
00:02:24,750 --> 00:02:28,260
puts us in a much more realistic
posture of what it's actually

47
00:02:28,260 --> 00:02:33,510
going to be like when robotics
start to scale up to a greater

48
00:02:33,510 --> 00:02:35,970
extent than they have so far.

49
00:02:35,970 --> 00:02:41,640
And his work-- we read JCR
Licklider when we took up

50
00:02:41,640 --> 00:02:45,980
the book The Dream Machine,
his biography by Mitch Waldrop.

51
00:02:45,980 --> 00:02:48,330
And Licklider, as
you remember, gave us

52
00:02:48,330 --> 00:02:50,970
this picture of how
people and machines were

53
00:02:50,970 --> 00:02:52,065
going to work together.

54
00:02:52,065 --> 00:02:54,450
And Licklider, in a
way, was responding

55
00:02:54,450 --> 00:02:57,505
to a period of
technological concern

56
00:02:57,505 --> 00:02:59,880
that computers were going to
replace people, because they

57
00:02:59,880 --> 00:03:01,980
could think better than people.

58
00:03:01,980 --> 00:03:05,880
And Licklider, although
a great technologist,

59
00:03:05,880 --> 00:03:08,550
was trained as a psychologist.

60
00:03:08,550 --> 00:03:12,090
And his field was
man-machine interface--

61
00:03:12,090 --> 00:03:13,947
how do people work
with machines?

62
00:03:13,947 --> 00:03:15,280
How does that actually function?

63
00:03:15,280 --> 00:03:19,920
And so he comes to computing
with this vision of a symbiosis

64
00:03:19,920 --> 00:03:22,660
between, computers are going
to do what they're good at,

65
00:03:22,660 --> 00:03:24,660
and people are going to
do what they're good at,

66
00:03:24,660 --> 00:03:28,050
and we're going to have a
symbiosis between the two

67
00:03:28,050 --> 00:03:30,300
that's going to optimize
both territories.

68
00:03:30,300 --> 00:03:35,100
And indeed, that's exactly the
direction that computers go in.

69
00:03:35,100 --> 00:03:37,830
In other words, I'm not
spending sleepless nights

70
00:03:37,830 --> 00:03:41,760
wondering when this thing is
going to replace me, right?

71
00:03:41,760 --> 00:03:43,950
We thought it was.

72
00:03:43,950 --> 00:03:46,380
And Norbert Wiener
at MIT had invented

73
00:03:46,380 --> 00:03:49,470
the term cybernetics and wrote
the first book on the topic.

74
00:03:49,470 --> 00:03:51,065
But Wiener had a
very dark vision

75
00:03:51,065 --> 00:03:52,440
of what computing
was going to be

76
00:03:52,440 --> 00:03:55,470
for the future of
human thinking.

77
00:03:55,470 --> 00:03:57,480
It didn't turn out to happen.

78
00:03:57,480 --> 00:04:02,850
So Mindell is arguing that
even the most threatening

79
00:04:02,850 --> 00:04:06,980
technology, which
arguably is robotics,

80
00:04:06,980 --> 00:04:08,720
it's just not going
to work out that way.

81
00:04:08,720 --> 00:04:12,630
A symbiosis is going to
be what's achieved here.

82
00:04:16,628 --> 00:04:19,170
And he takes apart each one of
these territories in the book.

83
00:04:19,170 --> 00:04:20,462
I mean, I had you do the video.

84
00:04:20,462 --> 00:04:22,740
But it's a very rich and
very well-written book.

85
00:04:22,740 --> 00:04:24,690
It's a tremendously
fun read, and he

86
00:04:24,690 --> 00:04:27,420
tells a lot of great stories.

87
00:04:27,420 --> 00:04:31,050
But his vision is that there
can be a richer space when

88
00:04:31,050 --> 00:04:34,640
people and robots are joined.

89
00:04:34,640 --> 00:04:36,480
There's a ritual
understanding and ability

90
00:04:36,480 --> 00:04:38,250
to make perceptions
and judgments that

91
00:04:38,250 --> 00:04:40,230
occurs from that mix.

92
00:04:40,230 --> 00:04:45,180
And that turns out to be
exactly his experience undersea

93
00:04:45,180 --> 00:04:49,470
and doing space robotics,
that there's just

94
00:04:49,470 --> 00:04:53,430
a different kind of
environment that evolves here.

95
00:04:53,430 --> 00:04:56,700
And it's better than a
machine alone or a person

96
00:04:56,700 --> 00:04:59,310
alone, that there's
a new symbiosis that

97
00:04:59,310 --> 00:05:03,660
enhances both sides in a
way that we can draw on.

98
00:05:03,660 --> 00:05:07,500
And what he finds is that
when people, in effect,

99
00:05:07,500 --> 00:05:10,590
start to operate the robotics--

100
00:05:10,590 --> 00:05:12,090
and in the end,
in these systems,

101
00:05:12,090 --> 00:05:14,340
autonomous systems
don't work very well--

102
00:05:14,340 --> 00:05:19,810
people really need to be part
of the process with the robot--

103
00:05:19,810 --> 00:05:23,300
that the robot becomes an
extension of themselves.

104
00:05:23,300 --> 00:05:28,160
And that, in fact, all their
perceptions are in the robot.

105
00:05:28,160 --> 00:05:32,750
So even if the robot is
a mile deeper undersea

106
00:05:32,750 --> 00:05:36,680
or tens of thousands
of miles away in space

107
00:05:36,680 --> 00:05:40,970
running around the moon, that
the person operating the robot

108
00:05:40,970 --> 00:05:44,120
and cooperating with the
robot is in the robot.

109
00:05:44,120 --> 00:05:46,940
And the sensory systems that
are available to the robot

110
00:05:46,940 --> 00:05:48,100
really become theirs.

111
00:05:48,100 --> 00:05:53,060
In effect, we get extended
with a whole new kind of reach.

112
00:05:53,060 --> 00:05:57,710
And that's really his
vision of what occurs here.

113
00:05:57,710 --> 00:06:01,340
That's the UAV story, that's
the undersea exploration

114
00:06:01,340 --> 00:06:03,380
story, that's the space story.

115
00:06:03,380 --> 00:06:06,860
And, he argues, this is going
to be the driverless car

116
00:06:06,860 --> 00:06:09,160
story, too.

117
00:06:09,160 --> 00:06:11,770
So we have been proceeding--

118
00:06:11,770 --> 00:06:14,260
and he goes through this at
rich length in his book--

119
00:06:14,260 --> 00:06:16,750
we've been proceeding
under the assumption

120
00:06:16,750 --> 00:06:21,130
that we're going to get
replaced by driverless cars.

121
00:06:21,130 --> 00:06:25,180
But he argues, wait a minute, 40
years of experience of robotics

122
00:06:25,180 --> 00:06:28,240
says that's not what happens.

123
00:06:28,240 --> 00:06:31,480
Instead, this symbiosis
starts to occur, right?

124
00:06:31,480 --> 00:06:34,000
Because people are not
going to want to give up

125
00:06:34,000 --> 00:06:35,770
certain kinds of
level of controls

126
00:06:35,770 --> 00:06:37,783
and certain kinds
of decision-making.

127
00:06:37,783 --> 00:06:39,700
They're going to want
to be involved in those.

128
00:06:39,700 --> 00:06:42,940
So he argues, that just as
with undersea exploration

129
00:06:42,940 --> 00:06:46,840
and with the other robotics
fields that he's talking about,

130
00:06:46,840 --> 00:06:49,990
a richer driver experience
that involves much more

131
00:06:49,990 --> 00:06:51,940
engagement with the
surroundings, much

132
00:06:51,940 --> 00:06:54,760
greater information and
knowledge access, much

133
00:06:54,760 --> 00:06:59,320
better safety and much better
comfort, a lower stress

134
00:06:59,320 --> 00:07:04,570
environment, can occur in
a driverless car setting.

135
00:07:04,570 --> 00:07:07,020
That people will still
be in command of the car.

136
00:07:07,020 --> 00:07:10,630
They'll be able to cede a
lot of lower-end territory

137
00:07:10,630 --> 00:07:13,760
to the robotic
activities, but will

138
00:07:13,760 --> 00:07:15,550
retain certain kinds
of charge and be

139
00:07:15,550 --> 00:07:18,400
able to operate in a different
kind of level of perception.

140
00:07:18,400 --> 00:07:19,690
We'll see, you know?

141
00:07:19,690 --> 00:07:21,280
Time will tell here.

142
00:07:21,280 --> 00:07:23,080
The driverless car
movement has been

143
00:07:23,080 --> 00:07:25,322
pushed by a whole
community that assumes

144
00:07:25,322 --> 00:07:27,280
that he wants to get
people out of the driver's

145
00:07:27,280 --> 00:07:30,610
seat into the trunk,
or at least watching

146
00:07:30,610 --> 00:07:32,720
movies in the back seat.

147
00:07:32,720 --> 00:07:35,705
Mindell argues that's actually
probably not going to occur.

148
00:07:35,705 --> 00:07:38,080
And then he lays out a series
of technological challenges

149
00:07:38,080 --> 00:07:39,850
that have to happen.

150
00:07:39,850 --> 00:07:42,580
So LIDAR, which is the
fundamental radar system we're

151
00:07:42,580 --> 00:07:44,050
using for driverless cars--

152
00:07:44,050 --> 00:07:45,580
and David knows a
lot about radar,

153
00:07:45,580 --> 00:07:47,163
because that's what
his new company is

154
00:07:47,163 --> 00:07:48,710
organizing right now--

155
00:07:48,710 --> 00:07:53,830
LIDAR happens to be very
problematic in wet surfaces

156
00:07:53,830 --> 00:07:54,655
and snowy services.

157
00:07:57,370 --> 00:07:59,260
It fails, right?

158
00:07:59,260 --> 00:08:02,720
That's why Google has based its
driverless car operation out

159
00:08:02,720 --> 00:08:04,690
in sunny California, right?

160
00:08:04,690 --> 00:08:08,350
Uber is having much more
trouble in Pittsburgh.

161
00:08:08,350 --> 00:08:10,512
And I'm sure your
friends, Steph,

162
00:08:10,512 --> 00:08:12,220
are going to have
trouble when they worry

163
00:08:12,220 --> 00:08:15,010
about Boston in the wintertime
with their driverless car

164
00:08:15,010 --> 00:08:17,800
initiatives.

165
00:08:17,800 --> 00:08:20,390
It's highly problematic.

166
00:08:20,390 --> 00:08:26,590
You can get around that by using
ground-penetrating radar, which

167
00:08:26,590 --> 00:08:31,610
avoids the surface water problem
and surface slickness problem.

168
00:08:31,610 --> 00:08:36,090
But at the moment, that's
another $70,000 per car.

169
00:08:36,090 --> 00:08:40,070
So the prices will get
driven down the cost curve,

170
00:08:40,070 --> 00:08:41,679
and maybe we can do that.

171
00:08:41,679 --> 00:08:45,250
But these are significant
barriers here.

172
00:08:45,250 --> 00:08:48,640
I was at a presentation
recently by Gill Pratt and John

173
00:08:48,640 --> 00:08:51,370
Leonard, who are leading
the Toyota driverless car

174
00:08:51,370 --> 00:08:53,320
initiative.

175
00:08:53,320 --> 00:08:56,800
And it was an
interesting discussion.

176
00:08:56,800 --> 00:09:01,810
The leader of the discussion
said, well, you know,

177
00:09:01,810 --> 00:09:05,710
we're going to have driverless
cars within five years.

178
00:09:05,710 --> 00:09:09,970
And Gill Pratt,
who led the whole

179
00:09:09,970 --> 00:09:12,030
robotics effort in DARPA--

180
00:09:12,030 --> 00:09:13,780
and a former MIT
faculty member, one

181
00:09:13,780 --> 00:09:17,560
of the most respected experts
in robotics in the world,

182
00:09:17,560 --> 00:09:21,250
and named by Toyota
to lead their effort--

183
00:09:21,250 --> 00:09:24,730
Gill had this line
and he said, you know,

184
00:09:24,730 --> 00:09:29,110
people tend to
completely overestimate

185
00:09:29,110 --> 00:09:31,465
what we're going to accomplish
in the next five years

186
00:09:31,465 --> 00:09:33,340
and completely underestimate
what we're going

187
00:09:33,340 --> 00:09:35,710
to accomplish in the next 50.

188
00:09:35,710 --> 00:09:38,740
We're not good at
making that balance.

189
00:09:38,740 --> 00:09:40,540
And in effect what
he was saying is,

190
00:09:40,540 --> 00:09:43,840
the driverless car problem
is not a five-year problem.

191
00:09:43,840 --> 00:09:45,590
And then John Leonard,
who is running

192
00:09:45,590 --> 00:09:48,830
MIT's part of the
Toyota initiative, one

193
00:09:48,830 --> 00:09:54,700
of MIT's famous robotics
experts, he put up a video.

194
00:09:54,700 --> 00:10:00,130
And the video was him
taking his kids to school.

195
00:10:00,130 --> 00:10:02,140
And they drive down a
residential street-- one

196
00:10:02,140 --> 00:10:06,050
of these kind of urbanized
suburbs around here.

197
00:10:06,050 --> 00:10:07,300
And there's a T-intersection.

198
00:10:07,300 --> 00:10:09,568
So they leave the
residential street,

199
00:10:09,568 --> 00:10:11,860
and they have to make a left
turn to get to the school.

200
00:10:11,860 --> 00:10:15,410
The kids are
sitting in the back.

201
00:10:15,410 --> 00:10:18,260
And in the lane
immediately in front of him

202
00:10:18,260 --> 00:10:20,990
going in that direction,
going to the right,

203
00:10:20,990 --> 00:10:23,750
are cars that are
erratically coming across

204
00:10:23,750 --> 00:10:26,960
at about 40 to 50 miles an hour.

205
00:10:26,960 --> 00:10:29,120
And the lane that he
wants to get into,

206
00:10:29,120 --> 00:10:31,520
it's bumper to bumper,
completely stalled.

207
00:10:31,520 --> 00:10:32,820
And he's got to turn left.

208
00:10:35,400 --> 00:10:38,310
He finally is able to solve
the problem, he indicates,

209
00:10:38,310 --> 00:10:40,470
by rolling down
his window, making

210
00:10:40,470 --> 00:10:45,750
a judgment about what driver
might be the most sympathetic.

211
00:10:45,750 --> 00:10:48,640
Waving, establishing
an eye contact,

212
00:10:48,640 --> 00:10:50,460
and this driver
is willing to let

213
00:10:50,460 --> 00:10:52,590
him pull in in front of her.

214
00:10:52,590 --> 00:10:53,910
But he's then got to wait--

215
00:10:53,910 --> 00:10:55,868
she's got to be at the
right point in the line,

216
00:10:55,868 --> 00:10:58,230
and the erratic 50-mile-an-hour
cars have got to let up

217
00:10:58,230 --> 00:10:59,550
a little bit so
he can get across.

218
00:10:59,550 --> 00:11:00,883
He's finally able to achieve it.

219
00:11:00,883 --> 00:11:02,510
He does this every morning.

220
00:11:02,510 --> 00:11:06,360
He said, look, I can't
write the algorithm

221
00:11:06,360 --> 00:11:08,790
for the left-hand turn
into heavy traffic.

222
00:11:08,790 --> 00:11:10,437
I can't do it.

223
00:11:10,437 --> 00:11:12,270
I've been thinking about
it for a long time.

224
00:11:12,270 --> 00:11:15,300
That is at least
a 10-year problem.

225
00:11:15,300 --> 00:11:18,510
How is my driverless car
going to establish eye contact

226
00:11:18,510 --> 00:11:21,090
with the driver
in that far lane.

227
00:11:21,090 --> 00:11:23,333
How are they going to
signal out the window?

228
00:11:23,333 --> 00:11:25,500
How are they going to make
that personal connection?

229
00:11:25,500 --> 00:11:28,290
All of us, when we
drive all the time,

230
00:11:28,290 --> 00:11:31,680
are making constant judgments
about other drivers.

231
00:11:31,680 --> 00:11:34,680
John Leonard says, I
can't write the algorithms

232
00:11:34,680 --> 00:11:38,490
to evaluate the crazy
18-year-old hot rod

233
00:11:38,490 --> 00:11:44,820
driver versus me versus an
elderly 65-year-old driver.

234
00:11:44,820 --> 00:11:48,510
I can't make those
differentials in my algorithms

235
00:11:48,510 --> 00:11:50,170
to make this thing work.

236
00:11:50,170 --> 00:11:53,150
So the problem of driving
in an urbanized setting

237
00:11:53,150 --> 00:11:54,150
is going to be profound.

238
00:11:54,150 --> 00:11:55,858
He said, look, we can
probably figure out

239
00:11:55,858 --> 00:11:57,900
interstate driving-- the
variables are much more

240
00:11:57,900 --> 00:11:59,550
manageable.

241
00:11:59,550 --> 00:12:03,640
But urbanized setting, it's
really, really complicated.

242
00:12:03,640 --> 00:12:07,020
So I guess what
Mindell is telling us--

243
00:12:07,020 --> 00:12:10,280
and those are just examples
that are not in his book,

244
00:12:10,280 --> 00:12:13,800
but I'm pulling up
from recent experience.

245
00:12:13,800 --> 00:12:17,320
But Mindell is telling us
that this autonomy is not

246
00:12:17,320 --> 00:12:18,390
necessarily at hand.

247
00:12:18,390 --> 00:12:22,020
And even when it is
available, and he

248
00:12:22,020 --> 00:12:26,880
uses the example of
the moon landing,

249
00:12:26,880 --> 00:12:31,030
the astronauts preempt it,
because they see things, i.e.

250
00:12:31,030 --> 00:12:34,813
a small crater, that would
have prevented the landing

251
00:12:34,813 --> 00:12:35,730
in the right location.

252
00:12:35,730 --> 00:12:38,147
So they have to take over the
controls and move the thing.

253
00:12:38,147 --> 00:12:40,140
People are not going
to want to give up

254
00:12:40,140 --> 00:12:43,782
that level of involvement,
he argues, and shouldn't,

255
00:12:43,782 --> 00:12:45,240
because there's
some things they're

256
00:12:45,240 --> 00:12:48,400
going to be better at judging
than even the best machinery we

257
00:12:48,400 --> 00:12:50,310
can come up with.

258
00:12:50,310 --> 00:12:53,860
So there's an example
that David uses,

259
00:12:53,860 --> 00:12:57,450
which is that a combination of
a chess master and a computer

260
00:12:57,450 --> 00:13:01,250
always beats another
chess master alone

261
00:13:01,250 --> 00:13:02,930
or a computer alone.

262
00:13:02,930 --> 00:13:05,570
And so far, that's
our experience.

263
00:13:05,570 --> 00:13:07,610
Now, all of these things
could change over time,

264
00:13:07,610 --> 00:13:09,800
and technology will
definitely advance.

265
00:13:09,800 --> 00:13:13,310
But what David is arguing is
that this robotics revolution

266
00:13:13,310 --> 00:13:16,670
is not going to be a
revolution around autonomy.

267
00:13:16,670 --> 00:13:19,760
It's going to be a revolution
like this revolution

268
00:13:19,760 --> 00:13:21,367
around symbiosis.

269
00:13:21,367 --> 00:13:23,450
And I think that's a really
important thing for us

270
00:13:23,450 --> 00:13:27,140
to keep in mind as we
think about the speed which

271
00:13:27,140 --> 00:13:29,150
these changes are going
to start to happen.

272
00:13:29,150 --> 00:13:31,910
The complexity is really high.

273
00:13:31,910 --> 00:13:34,580
And the nature of the
changes themselves.

274
00:13:34,580 --> 00:13:40,640
And in David Mindell's world
of cobotics and assistive

275
00:13:40,640 --> 00:13:43,520
robotics, there
is much more room

276
00:13:43,520 --> 00:13:47,930
for people than a truly
totally autonomous system.

277
00:13:47,930 --> 00:13:50,630
And that presents-- again, back
to our education discussion,

278
00:13:50,630 --> 00:13:52,880
that's back to our training
discussion-- that presents

279
00:13:52,880 --> 00:13:57,440
a whole new set of ways of
thinking about what this

280
00:13:57,440 --> 00:14:00,860
is going to look like that we
probably ought to keep in mind.

281
00:14:00,860 --> 00:14:03,590
And it also means that the
timetable of these changes

282
00:14:03,590 --> 00:14:08,030
is not as abrupt as the
media has been telling us

283
00:14:08,030 --> 00:14:10,580
in the last couple of years.

284
00:14:10,580 --> 00:14:12,260
It's a more manageable--

285
00:14:12,260 --> 00:14:13,830
David Mindell would argue--

286
00:14:13,830 --> 00:14:17,500
and more gradual
kind of timetable.

287
00:14:17,500 --> 00:14:18,990
All right.

288
00:14:18,990 --> 00:14:21,740
Sanam, it's all yours.

289
00:14:21,740 --> 00:14:22,240
SANAM: Yeah.

290
00:14:22,240 --> 00:14:26,450
So I think that the question
about symbiosis between people

291
00:14:26,450 --> 00:14:28,590
and machines is really
interesting, especially

292
00:14:28,590 --> 00:14:33,870
considering the debate about
whether people and robots are

293
00:14:33,870 --> 00:14:37,200
a lot like substitute goods
or complementary goods.

294
00:14:37,200 --> 00:14:39,840
And I think that
the consensus here

295
00:14:39,840 --> 00:14:43,338
is that they are
equally complementary.

296
00:14:43,338 --> 00:14:45,630
But I think that also has
some interesting implications

297
00:14:45,630 --> 00:14:50,260
for what that could do
to the human environment.

298
00:14:50,260 --> 00:14:53,880
So even with goods
that are complementary,

299
00:14:53,880 --> 00:14:57,000
it's likely that
this will push out

300
00:14:57,000 --> 00:14:59,280
the demand for
the types of labor

301
00:14:59,280 --> 00:15:02,492
that is more well-versed
in high skill [INAUDIBLE]

302
00:15:02,492 --> 00:15:04,200
in working with these
types of technology

303
00:15:04,200 --> 00:15:08,280
and reduced demand
for labor that's less.

304
00:15:08,280 --> 00:15:10,708
So I think that there are
really important considerations

305
00:15:10,708 --> 00:15:12,375
about when you're
integrating technology

306
00:15:12,375 --> 00:15:14,400
into a human environment,
how it's going

307
00:15:14,400 --> 00:15:16,750
to affect that environment.

308
00:15:16,750 --> 00:15:21,660
So I think just a question
that a lot of you had was,

309
00:15:21,660 --> 00:15:24,480
when you are looking at
building these autonomous or

310
00:15:24,480 --> 00:15:27,540
semi-autonomous technologies
into an existing

311
00:15:27,540 --> 00:15:30,570
human environment, it's going
to require a lot of knowledge

312
00:15:30,570 --> 00:15:32,550
about that environment
in order to make

313
00:15:32,550 --> 00:15:34,830
these important
predictive models.

314
00:15:34,830 --> 00:15:38,022
So what are some of the biggest
challenges in doing that?

315
00:15:38,022 --> 00:15:41,958
And where do you see the
potential failings or pitfalls?

316
00:15:44,910 --> 00:15:47,450
CHLOE: So one of the things
I was interested in--

317
00:15:47,450 --> 00:15:50,510
I really agree with a lot
of the points that he makes.

318
00:15:50,510 --> 00:15:54,830
I think he lays out really
optimistic, but also

319
00:15:54,830 --> 00:15:58,280
appropriately realistic future
for our symbiotic relationships

320
00:15:58,280 --> 00:16:00,480
with robots in a lot
of different fields.

321
00:16:00,480 --> 00:16:04,460
But [INAUDIBLE]
your question, he

322
00:16:04,460 --> 00:16:07,820
raises the issue of the pilot,
the soldier, and the astronaut

323
00:16:07,820 --> 00:16:12,650
who are all, to some extent, a
little reticent to fully adopt

324
00:16:12,650 --> 00:16:16,220
complete autonomy in their
own respective fields.

325
00:16:16,220 --> 00:16:19,100
So I guess one of
the places where

326
00:16:19,100 --> 00:16:21,545
you could consider a potential
failure of that symbiosis

327
00:16:21,545 --> 00:16:28,790
is, is that hesitation to adopt
the fully-autonomous system

328
00:16:28,790 --> 00:16:31,940
something we should train
out of these humans?

329
00:16:31,940 --> 00:16:34,790
Or should we adapt our
technological development

330
00:16:34,790 --> 00:16:35,390
to match it?

331
00:16:35,390 --> 00:16:37,370
Like is that a failure
point or is that OK?

332
00:16:37,370 --> 00:16:41,930
MAX: It's possible that the want
or the desire to go away from

333
00:16:41,930 --> 00:16:45,860
autonomy-- maybe they have some
sort of instinctual realization

334
00:16:45,860 --> 00:16:49,740
that the computer cannot do
everything that they can.

335
00:16:49,740 --> 00:16:51,950
It's not as good at
pattern recognition,

336
00:16:51,950 --> 00:16:53,000
that kind of thing.

337
00:16:53,000 --> 00:16:54,770
So maybe it's not
a failure point.

338
00:16:54,770 --> 00:16:56,310
Maybe instead it's a--

339
00:16:56,310 --> 00:16:56,810
[SNEEZING]

340
00:16:56,810 --> 00:16:58,430
I'm not sure what
you would call it.

341
00:16:58,430 --> 00:16:58,915
CHLOE: Interesting.

342
00:16:58,915 --> 00:16:59,540
MAX: Bless you.

343
00:16:59,540 --> 00:17:01,380
CHLOE: Like an actual
benefit to that.

344
00:17:01,380 --> 00:17:05,460
MAX: Yeah, almost
like a warning system.

345
00:17:05,460 --> 00:17:06,025
Yeah?

346
00:17:06,025 --> 00:17:06,650
AUDIENCE: Yeah.

347
00:17:06,650 --> 00:17:09,440
And on top of that, I feel
like there's a whole--

348
00:17:09,440 --> 00:17:11,420
especially thinking
about cars, but I'm

349
00:17:11,420 --> 00:17:13,910
sure for pilots and
other people as well--

350
00:17:13,910 --> 00:17:17,600
there's a lot of culture
around being a driver

351
00:17:17,600 --> 00:17:19,040
or being a pilot.

352
00:17:19,040 --> 00:17:21,500
You can define
yourself as a pilot.

353
00:17:21,500 --> 00:17:24,230
And I think by taking
away any control

354
00:17:24,230 --> 00:17:27,140
that you have over an
aircraft or something, that

355
00:17:27,140 --> 00:17:29,030
can be something that
people don't want

356
00:17:29,030 --> 00:17:31,543
to do just because
people like to drive,

357
00:17:31,543 --> 00:17:32,960
they like the open
road, they want

358
00:17:32,960 --> 00:17:34,460
to feel like they're in control.

359
00:17:34,460 --> 00:17:39,630
And taking that away could just
be a cultural issue as well.

360
00:17:39,630 --> 00:17:43,190
RASHEED: I thought autopilot was
pretty well-accepted and kind

361
00:17:43,190 --> 00:17:45,080
of integrated by now.

362
00:17:45,080 --> 00:17:49,070
I know, especially for
long-distance flights,

363
00:17:49,070 --> 00:17:50,750
cross-Atlantic, I
definitely don't

364
00:17:50,750 --> 00:17:54,592
want the pilot of this
airplane reading and charting

365
00:17:54,592 --> 00:17:56,300
the navigation across
the Atlantic Ocean.

366
00:17:56,300 --> 00:17:57,260
Like, I'm good.

367
00:17:57,260 --> 00:18:01,812
We can definitely have
more integrated technology

368
00:18:01,812 --> 00:18:03,020
to allow machines to do that.

369
00:18:03,020 --> 00:18:06,052
But at least within
the pilot community,

370
00:18:06,052 --> 00:18:07,510
I thought this was
pretty accepted.

371
00:18:07,510 --> 00:18:09,093
And now with fighter
pilots and things

372
00:18:09,093 --> 00:18:10,790
like that, you're
going so fast, it

373
00:18:10,790 --> 00:18:13,190
doesn't make sense to
have a human interpret

374
00:18:13,190 --> 00:18:15,264
all these signals
that are coming in.

375
00:18:15,264 --> 00:18:15,943
CHLOE: Sorry.

376
00:18:15,943 --> 00:18:17,110
RASHEED: Yeah, no, go ahead.

377
00:18:17,110 --> 00:18:18,445
CHLOE: No, you are right.

378
00:18:18,445 --> 00:18:21,110
It is very widely
accepted and integrated.

379
00:18:21,110 --> 00:18:23,990
But also, I think you make
an interesting point in terms

380
00:18:23,990 --> 00:18:28,190
of how people define
themselves and that job

381
00:18:28,190 --> 00:18:31,940
being actually part
of their identity.

382
00:18:31,940 --> 00:18:34,603
Also, I think as we've
developed more and more complex

383
00:18:34,603 --> 00:18:36,020
machines and those
machines become

384
00:18:36,020 --> 00:18:39,920
integrated into our life,
we interact with each other

385
00:18:39,920 --> 00:18:41,280
through those machines.

386
00:18:41,280 --> 00:18:46,130
So even just taking the
example of air traffic.

387
00:18:46,130 --> 00:18:48,980
There's a very complex
communication network

388
00:18:48,980 --> 00:18:55,070
between air traffic controllers
and different pilots

389
00:18:55,070 --> 00:18:58,520
from commercial aviation
to general aviation.

390
00:18:58,520 --> 00:19:03,350
And even though that system has
a lot of autonomous components

391
00:19:03,350 --> 00:19:05,000
in it in today's
day in age, there's

392
00:19:05,000 --> 00:19:14,700
still generations of
protocol, both very officially

393
00:19:14,700 --> 00:19:18,470
delineated and also sort
of adopted as practice.

394
00:19:18,470 --> 00:19:21,320
And I think that the
humans who are willingly

395
00:19:21,320 --> 00:19:23,990
part of that system as
controllers and pilots

396
00:19:23,990 --> 00:19:25,300
know that very well.

397
00:19:25,300 --> 00:19:29,080
And so they know how to
interact with the other pilots,

398
00:19:29,080 --> 00:19:34,010
both in manners and how to make
well-informed, safe decisions

399
00:19:34,010 --> 00:19:35,850
based on what other
people are doing.

400
00:19:35,850 --> 00:19:37,490
So I think-- yeah.

401
00:19:37,490 --> 00:19:40,380
I think that even though they
might be flying, technically,

402
00:19:40,380 --> 00:19:44,660
a completely autonomous
plane, they still

403
00:19:44,660 --> 00:19:48,735
serve as the end-point of the
communication with other humans

404
00:19:48,735 --> 00:19:51,110
who are in charge of other
completely autonomous systems.

405
00:19:51,110 --> 00:19:52,693
WILLIAM BONVILLIAN:
And Chloe, I think

406
00:19:52,693 --> 00:19:55,280
you're absolutely
right on that from what

407
00:19:55,280 --> 00:19:58,340
I understand from David's
book and discussions with him.

408
00:19:58,340 --> 00:20:02,600
In addition, he gives
a remarkable example.

409
00:20:02,600 --> 00:20:05,810
It's a startling example
that starts the book off.

410
00:20:05,810 --> 00:20:09,110
When you get a chance, the
book is really worth reading.

411
00:20:09,110 --> 00:20:17,030
He tells the story of an Air
France crash over the Pacific

412
00:20:17,030 --> 00:20:18,217
15 years ago.

413
00:20:18,217 --> 00:20:19,800
Chloe, I see you're
smiling, so you're

414
00:20:19,800 --> 00:20:20,700
thinking about the story.

415
00:20:20,700 --> 00:20:21,560
CHLOE: Yeah, I was just
thinking about this.

416
00:20:21,560 --> 00:20:23,720
WILLIAM BONVILLIAN: And
you can extrapolate and see

417
00:20:23,720 --> 00:20:24,850
if I get the story right.

418
00:20:24,850 --> 00:20:27,740
But essentially, the
story that he tells

419
00:20:27,740 --> 00:20:31,790
is that there's a moment in the
flight they're on autopilot.

420
00:20:31,790 --> 00:20:33,680
As you point out,
Rasheed, it can

421
00:20:33,680 --> 00:20:37,520
be a very useful tool
for pilots and is

422
00:20:37,520 --> 00:20:42,210
used in many parts
of a flight pattern.

423
00:20:42,210 --> 00:20:45,710
And they're on
autopilot, and the plane

424
00:20:45,710 --> 00:20:49,970
enters an area of very
low temperatures and ice

425
00:20:49,970 --> 00:20:52,100
is forming on the wings.

426
00:20:52,100 --> 00:20:56,660
So the autopilot system default
mechanism, at that point,

427
00:20:56,660 --> 00:21:00,920
is to remove the
autopilot control

428
00:21:00,920 --> 00:21:04,500
and switch it to
the pilot control.

429
00:21:04,500 --> 00:21:07,920
And what has happened is that
the pilot of the aircraft

430
00:21:07,920 --> 00:21:10,160
is back in the restroom,
because they're on autopilot

431
00:21:10,160 --> 00:21:12,260
over the middle of the Pacific.

432
00:21:12,260 --> 00:21:15,630
And the engineer, the flight
engineer and the co-pilot,

433
00:21:15,630 --> 00:21:16,850
are at the controls.

434
00:21:16,850 --> 00:21:19,430
They're relaxing.

435
00:21:19,430 --> 00:21:22,190
And suddenly, they're in charge.

436
00:21:22,190 --> 00:21:25,340
And each has a
different reaction.

437
00:21:25,340 --> 00:21:29,840
And it ends up completely
destabilizing the aircraft

438
00:21:29,840 --> 00:21:35,330
and throwing it into a spin
right into the Pacific Ocean,

439
00:21:35,330 --> 00:21:37,850
where all the lives
and aircraft are lost.

440
00:21:37,850 --> 00:21:40,040
And they had the opposite--
they're not ready.

441
00:21:40,040 --> 00:21:45,650
So this moment of people being
ready to take control again

442
00:21:45,650 --> 00:21:48,560
after having been out
of control turns out

443
00:21:48,560 --> 00:21:53,390
to be a nightmare in all kinds
of other settings, right?

444
00:21:53,390 --> 00:21:55,130
It's very unmanageable.

445
00:21:55,130 --> 00:21:59,930
How do you keep the person who
needs to be in overall control

446
00:21:59,930 --> 00:22:01,775
completely in the game?

447
00:22:01,775 --> 00:22:03,650
And how do you keep them
completely involved?

448
00:22:03,650 --> 00:22:05,720
So that's one of the
most complicated design

449
00:22:05,720 --> 00:22:08,360
problems for autonomous
vehicles in general,

450
00:22:08,360 --> 00:22:10,877
and driverless
cars in particular.

451
00:22:10,877 --> 00:22:13,460
Since we're probably not going
to get to fully driverless cars

452
00:22:13,460 --> 00:22:17,337
soon, how is that
relationship going to work?

453
00:22:17,337 --> 00:22:18,920
And how are we going
to play that out?

454
00:22:18,920 --> 00:22:21,560
And how do you keep just
enough of a level involved?

455
00:22:21,560 --> 00:22:24,530
And interesting, Rasheed,
back to your point,

456
00:22:24,530 --> 00:22:28,000
the autopilot system in aircraft
has actually now evolved

457
00:22:28,000 --> 00:22:31,550
to the point where
they keep that balance

458
00:22:31,550 --> 00:22:34,190
right between the
machine and the person

459
00:22:34,190 --> 00:22:36,140
and the overall
person in control

460
00:22:36,140 --> 00:22:39,860
and the desire of the person
to retain the control.

461
00:22:39,860 --> 00:22:41,390
They figured out
what the signaling

462
00:22:41,390 --> 00:22:44,150
is so that that actually
works well on both sides.

463
00:22:44,150 --> 00:22:46,317
But that's a really complicated
balance to work out.

464
00:22:46,317 --> 00:22:48,150
Have I got that right,
Chloe, would you say?

465
00:22:48,150 --> 00:22:49,160
CHLOE: Yeah, yeah.

466
00:22:49,160 --> 00:22:51,577
WILLIAM BONVILLIAN: Anything
you have to say more on this?

467
00:22:51,577 --> 00:22:52,680
Are you studying this?

468
00:22:52,680 --> 00:22:55,610
CHLOE: Well, in AeroAstro
we do a little bit

469
00:22:55,610 --> 00:22:56,780
of failure investigation.

470
00:22:56,780 --> 00:22:59,780
But I also fly
personally, and I've

471
00:22:59,780 --> 00:23:03,990
had a lot of experiences that
validate that exact thing.

472
00:23:03,990 --> 00:23:10,710
So I used to fly a pretty
small four-seater aircraft.

473
00:23:10,710 --> 00:23:12,490
MAX: Cessna?

474
00:23:12,490 --> 00:23:14,390
CHLOE: Yeah, well,
a Columbia, which

475
00:23:14,390 --> 00:23:17,270
is basically the same size as
a Cessna, but a lot faster.

476
00:23:17,270 --> 00:23:20,510
And I used to fly
with my dad a lot.

477
00:23:20,510 --> 00:23:24,230
And one of the stories that
we would repeatedly live out

478
00:23:24,230 --> 00:23:27,260
when we were landing at a
new airport was that the air

479
00:23:27,260 --> 00:23:30,740
traffic controller would,
based on the size and type

480
00:23:30,740 --> 00:23:33,260
of our aircraft, would
vastly underestimate

481
00:23:33,260 --> 00:23:35,960
the speed at which we
were making our descent.

482
00:23:35,960 --> 00:23:37,970
And a full descent into
most of these airports

483
00:23:37,970 --> 00:23:39,770
is a multi-leg
type thing, but you

484
00:23:39,770 --> 00:23:42,810
start up to 20 minutes before
you're actually on the ground.

485
00:23:42,810 --> 00:23:43,940
So it's pretty complicated.

486
00:23:43,940 --> 00:23:46,320
There are other
airplanes in the air.

487
00:23:46,320 --> 00:23:48,500
You have to match their
speeds and rates of descent,

488
00:23:48,500 --> 00:23:51,710
so you don't have two people
colliding on the runway.

489
00:23:51,710 --> 00:23:56,440
And repeatedly, how fast we were
going would be underestimated.

490
00:23:56,440 --> 00:24:00,598
And we would be coming down
much more rapidly than the air

491
00:24:00,598 --> 00:24:01,890
traffic controller would think.

492
00:24:01,890 --> 00:24:07,010
And so we'd be going faster
than a much, much larger

493
00:24:07,010 --> 00:24:08,090
aircraft was.

494
00:24:08,090 --> 00:24:11,600
So that was just another
case of where establishing

495
00:24:11,600 --> 00:24:14,180
that human contact
with other people

496
00:24:14,180 --> 00:24:17,570
to let them know where you
were and with the people whose

497
00:24:17,570 --> 00:24:21,980
instrumentation was telling
them something different,

498
00:24:21,980 --> 00:24:25,203
it's important to have people
who know their own missions--

499
00:24:25,203 --> 00:24:26,870
WILLIAM BONVILLIAN:
Interesting example.

500
00:24:26,870 --> 00:24:29,050
CHLOE: --involved
in the process.

501
00:24:29,050 --> 00:24:32,090
MARTIN: I think there's also an
opportunity for using autonomy

502
00:24:32,090 --> 00:24:34,970
in an area where
there's human bias.

503
00:24:34,970 --> 00:24:38,540
There was a case study of
a Japanese flight, where

504
00:24:38,540 --> 00:24:42,608
the co-pilot knows that what the
actual pilot is doing is wrong.

505
00:24:42,608 --> 00:24:44,150
But because he's an
authority figure,

506
00:24:44,150 --> 00:24:45,890
he won't contradict them.

507
00:24:45,890 --> 00:24:49,130
So using some autonomous things
to say like, no, this is wrong,

508
00:24:49,130 --> 00:24:51,255
we're definitely doing
something wrong, or at least

509
00:24:51,255 --> 00:24:53,560
put some kind of flag is useful.

510
00:24:53,560 --> 00:24:56,000
In sci-fi, it was
really cool, because it

511
00:24:56,000 --> 00:24:58,220
talked about human
bias injustice,

512
00:24:58,220 --> 00:24:59,842
like when you make a decision.

513
00:24:59,842 --> 00:25:01,550
And so there was this
one science fiction

514
00:25:01,550 --> 00:25:05,360
where in the future,
an AI was the judge,

515
00:25:05,360 --> 00:25:07,308
because he'd be non-biased.

516
00:25:07,308 --> 00:25:07,850
I don't know.

517
00:25:07,850 --> 00:25:09,658
I think that stuff's
pretty interesting.

518
00:25:09,658 --> 00:25:11,450
AUDIENCE: That's already
kind of happening,

519
00:25:11,450 --> 00:25:13,130
that they've been
using algorithms

520
00:25:13,130 --> 00:25:16,550
in some jurisdictions
to determine

521
00:25:16,550 --> 00:25:19,040
whether someone should
get bail or not based

522
00:25:19,040 --> 00:25:21,505
on their backgrounds.

523
00:25:21,505 --> 00:25:22,880
It's like a
proprietary software,

524
00:25:22,880 --> 00:25:24,050
which is kind of
scary to begin with,

525
00:25:24,050 --> 00:25:25,400
because people don't
really understand--

526
00:25:25,400 --> 00:25:25,640
MARTIN: How it works.

527
00:25:25,640 --> 00:25:27,390
AUDIENCE: --how it's
making its decisions.

528
00:25:27,390 --> 00:25:30,070
But you'll put in different
factors about the person,

529
00:25:30,070 --> 00:25:32,570
and that can help inform the
judge as to whether they should

530
00:25:32,570 --> 00:25:34,320
be let out on bail or not.

531
00:25:34,320 --> 00:25:35,990
And so I think there's a big--

532
00:25:35,990 --> 00:25:38,633
someone just filed a
lawsuit about that,

533
00:25:38,633 --> 00:25:40,550
since you're not allowed
to face your accuser,

534
00:25:40,550 --> 00:25:42,860
if your accuser is an algorithm.

535
00:25:42,860 --> 00:25:44,420
So we're already
getting into this--

536
00:25:44,420 --> 00:25:46,600
[INTERPOSING VOICES]

537
00:25:49,066 --> 00:25:51,470
RASHEED: Well, why would you
even start down that road?

538
00:25:51,470 --> 00:25:53,382
AUDIENCE: Well, because--
the justification--

539
00:25:53,382 --> 00:25:54,340
MARTIN: It's more fair.

540
00:25:54,340 --> 00:25:56,110
AUDIENCE: --whatever
you think of it--

541
00:25:56,110 --> 00:25:56,350
MAX: And it's faster.

542
00:25:56,350 --> 00:25:58,558
AUDIENCE: --is that there
are a lot of judges who are

543
00:25:58,558 --> 00:26:01,210
seeing hundreds of cases a day.

544
00:26:01,210 --> 00:26:03,395
A lot of things are affecting
their judgment from--

545
00:26:03,395 --> 00:26:05,770
I think the paper I read said
it could be like a therapy,

546
00:26:05,770 --> 00:26:07,660
or a team won the night
before, or they're

547
00:26:07,660 --> 00:26:11,700
hungry-- like they're influenced
by these outside factors.

548
00:26:11,700 --> 00:26:13,420
So if we automate
it, then we take

549
00:26:13,420 --> 00:26:16,020
out some of those
biases and other biases

550
00:26:16,020 --> 00:26:19,650
that judges just have inherently
from whatever their upbringing.

551
00:26:19,650 --> 00:26:21,487
RASHEED: Or like
compassion maybe?

552
00:26:21,487 --> 00:26:23,070
I don't know if
there's a compassion--

553
00:26:23,070 --> 00:26:24,460
[INTERPOSING VOICES]

554
00:26:25,440 --> 00:26:27,330
AUDIENCE: It's a very hard--

555
00:26:27,330 --> 00:26:29,788
MARTIN: I mean,
also the way I would

556
00:26:29,788 --> 00:26:31,330
think about it is,
great, because you

557
00:26:31,330 --> 00:26:32,320
get a nice heuristic.

558
00:26:32,320 --> 00:26:33,880
And if there's ever
an issue, that's

559
00:26:33,880 --> 00:26:37,013
the one that goes to court
and will get a full thing.

560
00:26:37,013 --> 00:26:38,680
Because there's
probably a lot of issues

561
00:26:38,680 --> 00:26:41,260
that are very quick cases.

562
00:26:41,260 --> 00:26:43,270
But there's a whole
thing at MIT about this

563
00:26:43,270 --> 00:26:46,510
now that they're doing,
like lawyers and tech.

564
00:26:46,510 --> 00:26:50,650
And it's around Sloan
and around the media lab.

565
00:26:50,650 --> 00:26:54,828
But it's a field that
MIT is looking into,

566
00:26:54,828 --> 00:26:56,995
but I don't know if we'll
come up with a law school.

567
00:26:56,995 --> 00:26:58,662
STEPH: But there was
a really cool study

568
00:26:58,662 --> 00:26:59,980
that I just read for my paper.

569
00:26:59,980 --> 00:27:00,760
I pulled it up.

570
00:27:00,760 --> 00:27:04,760
It got published literally
this week by Carnegie Mellon,

571
00:27:04,760 --> 00:27:06,850
entitled "A Human-Centered
Approach to Algorithmic

572
00:27:06,850 --> 00:27:08,980
Services,
Considerations for Fair

573
00:27:08,980 --> 00:27:11,182
and Motivating Smart
Community Service Management

574
00:27:11,182 --> 00:27:13,390
that Allocates Donations to
Nonprofit Organizations."

575
00:27:13,390 --> 00:27:14,350
So it's a very long title.

576
00:27:14,350 --> 00:27:14,770
But it's essentially--

577
00:27:14,770 --> 00:27:15,450
[LAUGHTER]

578
00:27:15,450 --> 00:27:17,383
WILLIAM BONVILLIAN: Yes, it is.

579
00:27:17,383 --> 00:27:20,050
The title of your paper is going
to be shorter, I'm sure, Steph.

580
00:27:20,050 --> 00:27:21,490
STEPH: I hope so.

581
00:27:21,490 --> 00:27:22,490
So there's three people.

582
00:27:22,490 --> 00:27:24,220
One of them is from the
Center for Machine Learning,

583
00:27:24,220 --> 00:27:25,887
another one's from
the School of Design,

584
00:27:25,887 --> 00:27:28,690
and then there's a stakeholder
from an organization called

585
00:27:28,690 --> 00:27:33,190
Food Rescue, which serves
the greater Pittsburgh area.

586
00:27:33,190 --> 00:27:36,550
And what was really cool
about this was essentially

587
00:27:36,550 --> 00:27:39,340
that they not only talked
to computer scientists who

588
00:27:39,340 --> 00:27:43,040
are designing the algorithms for
who gets the food allocations,

589
00:27:43,040 --> 00:27:46,420
but then they also
evaluated, I guess,

590
00:27:46,420 --> 00:27:49,720
the ethos that was driving the
way that the programmers were

591
00:27:49,720 --> 00:27:51,490
devising their algorithm
and the decisions

592
00:27:51,490 --> 00:27:53,830
that they had made, like
the cost-benefit analysis

593
00:27:53,830 --> 00:27:56,792
and who actually gets the
food, why they should get them,

594
00:27:56,792 --> 00:27:58,750
what considerations their
children should have,

595
00:27:58,750 --> 00:27:59,385
et cetera.

596
00:27:59,385 --> 00:28:00,760
And then from
there, they started

597
00:28:00,760 --> 00:28:04,210
thinking about the ways in which
the algorithm was designed,

598
00:28:04,210 --> 00:28:06,280
how that would impact,
in implementation,

599
00:28:06,280 --> 00:28:08,050
the lives of the people
receiving the food

600
00:28:08,050 --> 00:28:09,700
and whether or
not that was fair.

601
00:28:09,700 --> 00:28:11,410
And interestingly
enough, and I think

602
00:28:11,410 --> 00:28:14,620
that this is really, really
crucial in not only the design

603
00:28:14,620 --> 00:28:17,265
process, but generally in
policymaking, was the question

604
00:28:17,265 --> 00:28:18,640
that they bring
up about empathy.

605
00:28:18,640 --> 00:28:21,460
Like, to what extent
are the algorithms that

606
00:28:21,460 --> 00:28:25,000
are being created reflective
of the individual programmers'

607
00:28:25,000 --> 00:28:27,700
ethics and personal values?

608
00:28:27,700 --> 00:28:30,430
And that was something that
they really bring to light.

609
00:28:30,430 --> 00:28:33,100
And this, I think, is an
incredibly innovative case

610
00:28:33,100 --> 00:28:36,430
study, and it's being applied
to the nonprofit sector

611
00:28:36,430 --> 00:28:38,850
in a really small, piloted way.

612
00:28:38,850 --> 00:28:40,930
But I could see studies
like this scaling

613
00:28:40,930 --> 00:28:42,535
up to other
algorithms and really

614
00:28:42,535 --> 00:28:44,410
striving to understand
the ways in which they

615
00:28:44,410 --> 00:28:47,830
prejudice decision-making,
even if they act as heuristics.

616
00:28:47,830 --> 00:28:50,950
Heuristics are just utilizing
values and shortcuts

617
00:28:50,950 --> 00:28:53,680
to decision-making
based on your intuition

618
00:28:53,680 --> 00:28:55,420
in the decision-making process.

619
00:28:55,420 --> 00:28:56,963
So that could be really cool.

620
00:28:56,963 --> 00:28:58,630
But at the same time,
for me, as someone

621
00:28:58,630 --> 00:29:02,710
who studies both design and
politics, there's an enormous--

622
00:29:02,710 --> 00:29:06,410
and I guess my religious studies
also influence this a bit--

623
00:29:06,410 --> 00:29:08,740
there's a huge movement
in religious studies

624
00:29:08,740 --> 00:29:11,290
to move away from empathy,
because it can be utilized

625
00:29:11,290 --> 00:29:12,700
as a tool for manipulation.

626
00:29:12,700 --> 00:29:15,280
Because if you have insight
into what an individual matters

627
00:29:15,280 --> 00:29:18,160
and what motivates them to
action, then that is not good.

628
00:29:18,160 --> 00:29:19,870
But what, indeed,
you should have,

629
00:29:19,870 --> 00:29:22,490
as was stated by a
divinity professor at Yale,

630
00:29:22,490 --> 00:29:24,723
and I think he also does
some work in philosophy,

631
00:29:24,723 --> 00:29:26,140
is that what you
should strive for

632
00:29:26,140 --> 00:29:27,973
and what I think the
algorithm should strive

633
00:29:27,973 --> 00:29:31,360
for is compassion, not empathy,
which is distinct from empathy.

634
00:29:31,360 --> 00:29:33,250
The difference is
that in empathy, you

635
00:29:33,250 --> 00:29:35,910
seek to feel the way that
the other person feels.

636
00:29:35,910 --> 00:29:39,130
Where in compassion, you seek
to understand how they feel,

637
00:29:39,130 --> 00:29:41,920
but you don't truly try
to access those emotions.

638
00:29:41,920 --> 00:29:45,160
And so I think that
compassion, intellectually,

639
00:29:45,160 --> 00:29:48,100
is more sustainable and is
something that you can actively

640
00:29:48,100 --> 00:29:49,660
implement in an algorithm.

641
00:29:49,660 --> 00:29:52,730
Whereas if you try to make
your algorithm empathetic,

642
00:29:52,730 --> 00:29:56,650
it starts getting into the realm
of making valued judgments,

643
00:29:56,650 --> 00:29:59,050
which is ultimately going
to be damaging and more

644
00:29:59,050 --> 00:30:02,770
prejudicial than if an
individual is making

645
00:30:02,770 --> 00:30:03,310
a decision.

646
00:30:03,310 --> 00:30:07,390
Because then you can leave
the perjury to the machine

647
00:30:07,390 --> 00:30:10,720
and not place the impetus
of whatever discrimination

648
00:30:10,720 --> 00:30:14,485
is happening on an individual.

649
00:30:14,485 --> 00:30:16,740
MARTIN: Also, hacks.

650
00:30:16,740 --> 00:30:17,540
LILY: Always hacks.

651
00:30:17,540 --> 00:30:18,207
RASHEED: Thanks.

652
00:30:18,207 --> 00:30:21,757
[INTERPOSING VOICES]

653
00:30:21,757 --> 00:30:24,090
STEPH: Yeah, the Microsoft--
just quickly on that point.

654
00:30:24,090 --> 00:30:24,390
LILY: Mm-hmm.

655
00:30:24,390 --> 00:30:26,400
STEPH: They just released an
article in The New York Times,

656
00:30:26,400 --> 00:30:28,530
I think it was yesterday,
on the hack that

657
00:30:28,530 --> 00:30:29,970
happened in England on the NHS.

658
00:30:29,970 --> 00:30:34,440
And they were wondering,
what is the role

659
00:30:34,440 --> 00:30:36,450
of Microsoft in all of this?

660
00:30:36,450 --> 00:30:37,836
Should they be to blame?

661
00:30:37,836 --> 00:30:40,360
So, a question.

662
00:30:40,360 --> 00:30:43,060
LILY: Well, the whole time I
was watching this, Mindell,

663
00:30:43,060 --> 00:30:44,420
he's brilliant, obviously.

664
00:30:44,420 --> 00:30:46,288
But I also think that he--

665
00:30:46,288 --> 00:30:47,830
I couldn't help but
think that he was

666
00:30:47,830 --> 00:30:49,090
being a little hypocritical.

667
00:30:49,090 --> 00:30:52,890
Because I think his whole
message and premise was

668
00:30:52,890 --> 00:30:55,510
don't worry about
job displacement

669
00:30:55,510 --> 00:30:59,530
by autonomy or robotics,
autonomous robots.

670
00:30:59,530 --> 00:31:02,050
It'll never happen, because
the symbiosis of humans

671
00:31:02,050 --> 00:31:04,830
and machines is more powerful
than the machines alone

672
00:31:04,830 --> 00:31:06,190
or the humans alone.

673
00:31:06,190 --> 00:31:11,110
But I don't think that we can
deny that certain jobs have

674
00:31:11,110 --> 00:31:14,890
been displaced or replaced
by automated machines,

675
00:31:14,890 --> 00:31:16,150
robotics, et cetera.

676
00:31:16,150 --> 00:31:20,215
And although more jobs can
be created in the future

677
00:31:20,215 --> 00:31:24,430
or as a result of
improved technologies,

678
00:31:24,430 --> 00:31:27,040
we can't really get
around the fact, I think,

679
00:31:27,040 --> 00:31:31,183
that the people who do lose
those jobs often have families.

680
00:31:31,183 --> 00:31:33,100
And then, as we've been
talking earlier today,

681
00:31:33,100 --> 00:31:34,308
then they're out of the game.

682
00:31:34,308 --> 00:31:36,220
They don't go back
for a four-year--

683
00:31:36,220 --> 00:31:37,810
an advanced degree--
because, oh,

684
00:31:37,810 --> 00:31:41,860
now I need a PhD or
a master's degree.

685
00:31:41,860 --> 00:31:47,110
They're often permanently
unemployed or displaced

686
00:31:47,110 --> 00:31:49,330
into other service sectors.

687
00:31:49,330 --> 00:31:52,618
So, yeah, David
Mindell, you're right.

688
00:31:52,618 --> 00:31:54,910
We're probably never going
to be completely autonomous.

689
00:31:54,910 --> 00:31:56,539
But then again, there
are jobs-- like,

690
00:31:56,539 --> 00:31:57,956
you can't deny the
fact that there

691
00:31:57,956 --> 00:31:59,081
are jobs that are replaced.

692
00:31:59,081 --> 00:32:01,900
CHLOE: I had the
exact same feeling.

693
00:32:01,900 --> 00:32:06,100
I wouldn't necessarily say
hypocritical, but maybe biased.

694
00:32:06,100 --> 00:32:07,850
LILY: He's never going
to be out of a job.

695
00:32:07,850 --> 00:32:09,100
He's the person
who invents these.

696
00:32:09,100 --> 00:32:10,930
Like, yeah, of course,
he's the one who's

697
00:32:10,930 --> 00:32:12,910
staying ahead of the
technological race,

698
00:32:12,910 --> 00:32:14,960
because he is the
technological race, you know?

699
00:32:14,960 --> 00:32:15,460
[LAUGHS]

700
00:32:15,460 --> 00:32:18,490
CHLOE: But the symbiosis is
very real for the scientists

701
00:32:18,490 --> 00:32:22,180
and engineers whose
limits have been extended

702
00:32:22,180 --> 00:32:24,130
by these types of endeavors.

703
00:32:24,130 --> 00:32:25,505
But for the people
who have never

704
00:32:25,505 --> 00:32:26,880
gotten to that
level of education

705
00:32:26,880 --> 00:32:28,653
in the first place,
yeah, [INAUDIBLE]..

706
00:32:28,653 --> 00:32:30,815
MAX: The thing
is, he pointed out

707
00:32:30,815 --> 00:32:33,190
like, OK, yes, there are people
in manufacturing jobs who

708
00:32:33,190 --> 00:32:33,970
get replaced.

709
00:32:33,970 --> 00:32:35,800
But it's similar to how--

710
00:32:35,800 --> 00:32:38,175
I think it was one of the
other readings, where they were

711
00:32:38,175 --> 00:32:41,050
saying that people who
would have to light lamps

712
00:32:41,050 --> 00:32:43,510
in the street got replaced
by electrical lights,

713
00:32:43,510 --> 00:32:47,080
or people who would shout
news to factory workers

714
00:32:47,080 --> 00:32:48,380
were replaced by radios.

715
00:32:48,380 --> 00:32:49,865
You know, it's going to happen.

716
00:32:49,865 --> 00:32:51,240
LILY: Yeah, it's
going to happen.

717
00:32:51,240 --> 00:32:52,365
But I think he's taking a--

718
00:32:52,365 --> 00:32:53,740
MAX: It sucks, but it happens.

719
00:32:53,740 --> 00:32:54,823
That's how progress works.

720
00:32:54,823 --> 00:32:57,340
LILY: --a self-centric view.

721
00:32:57,340 --> 00:32:59,830
MAX: I don't know really what
to-- what's the alternative,

722
00:32:59,830 --> 00:33:00,850
stop progressing?

723
00:33:00,850 --> 00:33:03,400
Because then other
countries start progressing,

724
00:33:03,400 --> 00:33:04,780
and then you're left behind.

725
00:33:04,780 --> 00:33:07,750
And then everyone's unemployed,
because now your country

726
00:33:07,750 --> 00:33:08,613
doesn't have money.

727
00:33:08,613 --> 00:33:09,280
That's not good.

728
00:33:09,280 --> 00:33:11,020
MARTIN: Yeah, [INAUDIBLE]
or somebody else will.

729
00:33:11,020 --> 00:33:12,437
STEPH: [INAUDIBLE],,
I just wanted

730
00:33:12,437 --> 00:33:15,028
to make a quick point about
the concept of fetishization.

731
00:33:15,028 --> 00:33:15,820
MAX: Fetishization?

732
00:33:15,820 --> 00:33:16,040
STEPH: Fetishization, yeah.

733
00:33:16,040 --> 00:33:18,890
MARTIN: Technology fetishization
or a product fetishization?

734
00:33:18,890 --> 00:33:22,600
STEPH: Yeah, because I think,
at least in my understanding,

735
00:33:22,600 --> 00:33:24,820
fetishization in
post-modern philosophy

736
00:33:24,820 --> 00:33:27,340
is the process by which you
love the idea of something

737
00:33:27,340 --> 00:33:29,260
so much that you
relentlessly pursue it.

738
00:33:29,260 --> 00:33:32,570
And then you continue operating
in the pursuit of that, right?

739
00:33:32,570 --> 00:33:34,600
And the question that--

740
00:33:34,600 --> 00:33:38,210
as much as I love technology
and as I'm fascinated by it,

741
00:33:38,210 --> 00:33:42,460
and as much as I someday
maybe want to be an engineer,

742
00:33:42,460 --> 00:33:45,250
I ask myself, how
is it that we have

743
00:33:45,250 --> 00:33:49,270
come so far in autonomous
vehicular technology,

744
00:33:49,270 --> 00:33:53,130
and we can't get more
public bus routes?

745
00:33:53,130 --> 00:33:55,380
It doesn't seem like a
trade-off to people, but to me,

746
00:33:55,380 --> 00:33:56,890
that's an enormous concern.

747
00:33:56,890 --> 00:33:59,560
And I'm constantly looking up
questions like this on Quora.

748
00:33:59,560 --> 00:34:00,910
And people are like,
well, it doesn't matter,

749
00:34:00,910 --> 00:34:02,870
because they don't have
to be mutually exclusive.

750
00:34:02,870 --> 00:34:04,260
WILLIAM BONVILLIAN: Well,
this is the whole problem

751
00:34:04,260 --> 00:34:06,140
with change in legacy sectors.

752
00:34:06,140 --> 00:34:08,639
How do you introduce innovation
in the legacy sectors?

753
00:34:08,639 --> 00:34:10,929
That's at the heart of
that kind of question.

754
00:34:10,929 --> 00:34:15,670
And changing a public
transportation system,

755
00:34:15,670 --> 00:34:18,699
an established sector,
may, in many ways,

756
00:34:18,699 --> 00:34:21,940
be far harder than introducing
the completely disruptive

757
00:34:21,940 --> 00:34:24,909
technology that's
completely outside the scope

758
00:34:24,909 --> 00:34:26,453
of an existing realm.

759
00:34:26,453 --> 00:34:28,120
MAX: Also, doesn't
this fit in with what

760
00:34:28,120 --> 00:34:31,918
Bill was saying before, where
city driving is unbelievably

761
00:34:31,918 --> 00:34:33,460
difficult, and we
probably will never

762
00:34:33,460 --> 00:34:35,400
get autonomous city driving.

763
00:34:35,400 --> 00:34:37,559
WILLIAM BONVILLIAN: Well,
I'm not saying never.

764
00:34:37,559 --> 00:34:38,770
I'm just saying it's--

765
00:34:38,770 --> 00:34:41,030
MAX: Well, it'll take
a really long time.

766
00:34:41,030 --> 00:34:43,530
WILLIAM BONVILLIAN: This is
more than a decade of a problem.

767
00:34:43,530 --> 00:34:49,793
STEPH: But I guess I'm
curious, why do we view these--

768
00:34:49,793 --> 00:34:51,460
I'm not going to say
nearly impossible--

769
00:34:51,460 --> 00:34:56,199
but these incredibly technically
challenging problems as more

770
00:34:56,199 --> 00:34:59,620
solvable than getting more
public buses on the route?

771
00:34:59,620 --> 00:35:02,380
Why do we have more persistence
and resilience when it

772
00:35:02,380 --> 00:35:03,010
comes to technical innovation?

773
00:35:03,010 --> 00:35:04,750
WILLIAM BONVILLIAN: Because
it's a frontier territory,

774
00:35:04,750 --> 00:35:06,580
and the barriers
aren't in the way.

775
00:35:06,580 --> 00:35:07,630
MARTIN: Yeah, there's
a lot more barriers

776
00:35:07,630 --> 00:35:09,150
for making a train
and how much it costs.

777
00:35:09,150 --> 00:35:09,510
WILLIAM BONVILLIAN: Right.

778
00:35:09,510 --> 00:35:11,177
STEPH: But I'm not
talking about trains.

779
00:35:11,177 --> 00:35:12,842
Literally, public buses.

780
00:35:12,842 --> 00:35:14,800
RASHEED: No, I think you
were trying to get at,

781
00:35:14,800 --> 00:35:18,055
it's like there's no
barrier to figuring out

782
00:35:18,055 --> 00:35:20,680
what the new autonomous vehicle
is, because there's no existing

783
00:35:20,680 --> 00:35:22,222
infrastructure on
autonomous vehicles

784
00:35:22,222 --> 00:35:23,650
that I have to overcome.

785
00:35:23,650 --> 00:35:26,260
Whereas in order to get
new public bus routes,

786
00:35:26,260 --> 00:35:29,170
I have to contend
with the established

787
00:35:29,170 --> 00:35:31,760
system of public bus routes
and kind of go through that.

788
00:35:31,760 --> 00:35:33,230
WILLIAM BONVILLIAN: Interesting.

789
00:35:33,230 --> 00:35:35,230
MARTIN: There was a
start-up that was like Uber,

790
00:35:35,230 --> 00:35:37,853
but for buses, which is
optimized for people's-- huh?

791
00:35:37,853 --> 00:35:38,520
AUDIENCE: BRIDJ?

792
00:35:38,520 --> 00:35:38,930
MARTIN: Something like that.

793
00:35:38,930 --> 00:35:40,555
AUDIENCE: I think
they just went under.

794
00:35:40,555 --> 00:35:43,930
MARTIN: Yeah, it's just
a lot of the financing--

795
00:35:43,930 --> 00:35:48,270
the economic cycle doesn't
churn continuously.

796
00:35:48,270 --> 00:35:49,390
It needs to get spun.

797
00:35:49,390 --> 00:35:50,370
But a quick point--

798
00:35:50,370 --> 00:35:51,470
I think you've brought
a really good point

799
00:35:51,470 --> 00:35:53,720
in terms of the people
writing these papers.

800
00:35:53,720 --> 00:35:56,670
I was at a talk by a
diplomat on Trump in Mexico.

801
00:35:56,670 --> 00:35:58,045
And he had this
really funny term

802
00:35:58,045 --> 00:36:00,400
where he's like, yeah, it's
kind of interesting to read

803
00:36:00,400 --> 00:36:02,317
a lot of these papers,
because a lot of them--

804
00:36:02,317 --> 00:36:04,090
or like when tech
billionaires talk about,

805
00:36:04,090 --> 00:36:06,640
yeah, we're going to do this
and help the [INAUDIBLE] go

806
00:36:06,640 --> 00:36:08,500
to their dinner, he's
like, yeah, they're

807
00:36:08,500 --> 00:36:10,600
kind of like limousine
liberals, which

808
00:36:10,600 --> 00:36:12,035
I thought was a
really funny term,

809
00:36:12,035 --> 00:36:14,410
not to criticize liberals.

810
00:36:14,410 --> 00:36:17,020
But I think the people that
read 10,000-- there's also

811
00:36:17,020 --> 00:36:21,760
a Chinese saying, which is,
much better to walk 10,000 miles

812
00:36:21,760 --> 00:36:25,213
and talk to the people on the
way than to read 10,000 books,

813
00:36:25,213 --> 00:36:26,380
that I think is really true.

814
00:36:26,380 --> 00:36:27,330
Where it's like--

815
00:36:27,330 --> 00:36:28,760
WILLIAM BONVILLIAN: Are you
familiar with that one, Luyao?

816
00:36:28,760 --> 00:36:29,260
Good.

817
00:36:29,260 --> 00:36:30,625
[LAUGHTER]

818
00:36:30,625 --> 00:36:31,500
I wanted to certify--

819
00:36:31,500 --> 00:36:33,920
[INTERPOSING VOICES]

820
00:36:34,420 --> 00:36:35,850
LILY: Verified!

821
00:36:35,850 --> 00:36:36,790
WILLIAM BONVILLIAN: It's
a great line, Martin.

822
00:36:36,790 --> 00:36:39,332
I wanted to use it myself, but
I wanted to validate it first.

823
00:36:39,332 --> 00:36:39,890
[LAUGHTER]

824
00:36:39,890 --> 00:36:40,990
MARTIN: I liked the
limousine liberals.

825
00:36:40,990 --> 00:36:42,190
And I was like, yeah,
that was pretty funny.

826
00:36:42,190 --> 00:36:42,910
WILLIAM BONVILLIAN:
That's an old term.

827
00:36:42,910 --> 00:36:44,160
That term's been
around for [INAUDIBLE]..

828
00:36:44,160 --> 00:36:44,290
MARTIN: I didn't know.

829
00:36:44,290 --> 00:36:45,200
STEPH: I'd never
heard of that before,

830
00:36:45,200 --> 00:36:46,510
but it makes total sense.

831
00:36:46,510 --> 00:36:50,790
MARTIN: But it does really
classify the Silicon Valley

832
00:36:50,790 --> 00:36:52,180
kind of--

833
00:36:52,180 --> 00:36:53,120
STEPH: Libertarianism?

834
00:36:53,120 --> 00:36:54,470
MARTIN: Yeah.

835
00:36:54,470 --> 00:36:56,610
But I think the big
issue is we really

836
00:36:56,610 --> 00:36:59,140
need to start talking to
the people on the street

837
00:36:59,140 --> 00:37:01,390
and what they're thinking--

838
00:37:01,390 --> 00:37:04,385
what they think about the
issue, rather than just like,

839
00:37:04,385 --> 00:37:05,380
this is what could be.

840
00:37:05,380 --> 00:37:07,000
Also, put it into practice,
because in practice it could

841
00:37:07,000 --> 00:37:08,360
be a lot of different things.

842
00:37:08,360 --> 00:37:10,720
And most people just
want to do good.

843
00:37:10,720 --> 00:37:12,150
And sometimes their words--

844
00:37:12,150 --> 00:37:14,110
I'm clouded by the way
I view those words.

845
00:37:14,110 --> 00:37:16,120
You know what I mean?

846
00:37:16,120 --> 00:37:18,970
STEPH: I'm just really
perplexed by the fact

847
00:37:18,970 --> 00:37:22,220
that people will spend 50 years
trying to develop a technology.

848
00:37:22,220 --> 00:37:24,550
But in two months
or two years, you

849
00:37:24,550 --> 00:37:28,000
can't increase support
for public transit.

850
00:37:28,000 --> 00:37:30,625
That, I think, is to me, what's
insane about the public sector.

851
00:37:30,625 --> 00:37:32,083
AUDIENCE: Maybe
[INAUDIBLE] because

852
00:37:32,083 --> 00:37:33,250
of the nature of the good.

853
00:37:33,250 --> 00:37:35,050
One is public good
and the other one

854
00:37:35,050 --> 00:37:42,365
is marketable, tradable, private
goods, and there's profit.

855
00:37:42,365 --> 00:37:44,740
CHLOE: Also, to tie it back
to one of your earlier points

856
00:37:44,740 --> 00:37:48,280
about how we have these
Snapchat-esque companies,

857
00:37:48,280 --> 00:37:51,520
Facebook-esque companies,
before we have the big players.

858
00:37:51,520 --> 00:37:53,440
And also because
it's easier maybe now

859
00:37:53,440 --> 00:37:56,740
for a group of 20
to 30 scientists

860
00:37:56,740 --> 00:37:58,690
and engineers to
make a Mars rover

861
00:37:58,690 --> 00:38:01,620
and have that to be a
successful mission--

862
00:38:01,620 --> 00:38:04,500
to have a lot of little
things like that,

863
00:38:04,500 --> 00:38:07,730
where they can define how
they want their innovation

864
00:38:07,730 --> 00:38:12,180
to operate before we can then
take that infrastructure that

865
00:38:12,180 --> 00:38:14,740
is established by those
pioneers and apply it

866
00:38:14,740 --> 00:38:15,967
to public good work.

867
00:38:15,967 --> 00:38:17,800
WILLIAM BONVILLIAN: You
know, interestingly,

868
00:38:17,800 --> 00:38:21,700
Sanjay Sarma makes the
argument-- of MIT--

869
00:38:21,700 --> 00:38:25,900
he makes the argument
that we're not

870
00:38:25,900 --> 00:38:29,410
going to get to anything
resembling autonomous vehicles

871
00:38:29,410 --> 00:38:33,340
until we integrate the
systems we're developing

872
00:38:33,340 --> 00:38:34,690
into the infrastructure itself.

873
00:38:34,690 --> 00:38:39,490
In other words, suppose rather
than trying to make the vehicle

874
00:38:39,490 --> 00:38:42,640
entirely independent of
everything that's around it--

875
00:38:42,640 --> 00:38:45,730
people, everything
that's around--

876
00:38:45,730 --> 00:38:48,580
suppose we integrated the
vehicle into a new set

877
00:38:48,580 --> 00:38:52,340
of smart infrastructure.

878
00:38:52,340 --> 00:38:58,480
So in effect, there are
rails for every car, right?

879
00:38:58,480 --> 00:39:01,120
They just happen
to be cyber rails.

880
00:39:01,120 --> 00:39:05,750
And that would simplify all
kinds of things in the autonomy

881
00:39:05,750 --> 00:39:07,265
project.

882
00:39:07,265 --> 00:39:09,140
LILY: Or like parking
in a parking structure,

883
00:39:09,140 --> 00:39:12,897
where you just get to a dock
and the parking structure

884
00:39:12,897 --> 00:39:13,480
deals with it.

885
00:39:13,480 --> 00:39:13,546
WILLIAM BONVILLIAN: Right.

886
00:39:13,546 --> 00:39:15,520
So in other words,
making the infrastructure

887
00:39:15,520 --> 00:39:19,600
smart in parallel with
making the vehicle smart

888
00:39:19,600 --> 00:39:23,890
radically resolves a lot
of the problems here.

889
00:39:23,890 --> 00:39:25,810
And yet, we haven't
even thought about that.

890
00:39:25,810 --> 00:39:29,850
We're so busy embarked on making
the vehicle totally independent

891
00:39:29,850 --> 00:39:32,350
that we haven't even thought
about a much more logical path,

892
00:39:32,350 --> 00:39:35,262
he would argue,
which is to upgrade

893
00:39:35,262 --> 00:39:36,970
the infrastructure at
the same time we're

894
00:39:36,970 --> 00:39:38,390
upgrading the vehicles.

895
00:39:38,390 --> 00:39:42,520
And he argues, what's
going to happen to people?

896
00:39:42,520 --> 00:39:44,650
Well, you've got a
massive infrastructure

897
00:39:44,650 --> 00:39:48,400
upgrade, which is going to
be heavy employment focused.

898
00:39:48,400 --> 00:39:49,820
It's a significant
kind of offset.

899
00:39:49,820 --> 00:39:51,700
And we have a thought
about doing that.

900
00:39:51,700 --> 00:39:54,997
And that gets into, in a
way, some of the issues

901
00:39:54,997 --> 00:39:56,080
that you're talking about.

902
00:39:56,080 --> 00:39:56,470
STEPH: Yeah.

903
00:39:56,470 --> 00:39:58,780
And what's interesting is that
in the article that I conducted

904
00:39:58,780 --> 00:40:00,730
with the co-founder of
New Urban Mechanics,

905
00:40:00,730 --> 00:40:04,810
he said that the firms that
he has talked to, and NuTonomy

906
00:40:04,810 --> 00:40:07,270
specifically-- not
to, so to speak,

907
00:40:07,270 --> 00:40:08,623
throw them under the bus--

908
00:40:08,623 --> 00:40:10,540
they don't think that
those are considerations

909
00:40:10,540 --> 00:40:11,635
that fall in their domain.

910
00:40:11,635 --> 00:40:13,510
They think that the
market is, at some point,

911
00:40:13,510 --> 00:40:15,130
going to decide these
things for them,

912
00:40:15,130 --> 00:40:17,470
and so they should just worry
on fixing the technology.

913
00:40:17,470 --> 00:40:19,797
But if there's anything
that this class is teaching

914
00:40:19,797 --> 00:40:21,880
us is that we need to
include these considerations

915
00:40:21,880 --> 00:40:23,800
within the research and
development process,

916
00:40:23,800 --> 00:40:26,400
even at the basic
or early stages.

917
00:40:26,400 --> 00:40:27,670
WILLIAM BONVILLIAN: Right.

918
00:40:27,670 --> 00:40:30,610
So, Sanam, how about a closing
thought on David Mindell.

919
00:40:30,610 --> 00:40:32,120
We've had a robust discussion.

920
00:40:32,120 --> 00:40:33,460
Thank you all.

921
00:40:33,460 --> 00:40:35,410
I think everybody got
into it this time, too.

922
00:40:35,410 --> 00:40:36,910
MAX: I love the
under the bus thing.

923
00:40:36,910 --> 00:40:37,982
[LAUGHTER]

924
00:40:38,710 --> 00:40:40,960
SANAM: Yeah, I think it's
interesting that this debate

925
00:40:40,960 --> 00:40:44,260
about autonomous,
semi-autonomous technology

926
00:40:44,260 --> 00:40:46,923
raises some really
philosophical questions, first,

927
00:40:46,923 --> 00:40:48,340
about what we were
talking about--

928
00:40:48,340 --> 00:40:51,880
humans trust and agency
when it comes to technology.

929
00:40:51,880 --> 00:40:54,580
And then also about
what the goals

930
00:40:54,580 --> 00:40:57,490
should be of when we talk
about integrating technology

931
00:40:57,490 --> 00:41:02,620
into deeper and deeper parts of
life, what should the goals be?

932
00:41:02,620 --> 00:41:04,550
How much should there
be intervention on?

933
00:41:04,550 --> 00:41:07,300
And how could we improve
the overall infrastructure?

934
00:41:07,300 --> 00:41:12,715
So I think Mindell,
at some point,

935
00:41:12,715 --> 00:41:14,215
talked about how
this is engineering

936
00:41:14,215 --> 00:41:15,760
at its philosophical best.

937
00:41:15,760 --> 00:41:17,438
So I think that's an
interesting point.

938
00:41:17,438 --> 00:41:18,730
WILLIAM BONVILLIAN: Good point.

939
00:41:18,730 --> 00:41:20,880
Good closing points, Sanam.

940
00:41:20,880 --> 00:41:26,728
All right, one more to
go, and you're back to me.

941
00:41:26,728 --> 00:41:28,270
RASHEED: You saved
the best for last?

942
00:41:28,270 --> 00:41:30,190
[LAUGHTER]

943
00:41:30,190 --> 00:41:32,403
WILLIAM BONVILLIAN: You
could be the judge, Rasheed.

944
00:41:32,403 --> 00:41:33,820
MARTIN: That was
such a good line.

945
00:41:33,820 --> 00:41:34,816
[LAUGHTER]

946
00:41:34,816 --> 00:41:37,900
WILLIAM BONVILLIAN: But that
is the cover of my next book

947
00:41:37,900 --> 00:41:39,100
written with Peter Singer.

948
00:41:39,100 --> 00:41:40,400
These are both Peter
and I out hiking.

949
00:41:40,400 --> 00:41:40,976
MARTIN: Nice
pictures too, though.

950
00:41:40,976 --> 00:41:42,008
[LAUGHTER]

951
00:41:42,008 --> 00:41:44,550
RASHEED: So he took one of you,
and then you took one of him?

952
00:41:44,550 --> 00:41:46,470
WILLIAM BONVILLIAN: No, I think
his mother took that picture.

953
00:41:46,470 --> 00:41:46,870
MARTIN: Their new jackets.

954
00:41:46,870 --> 00:41:49,420
WILLIAM BONVILLIAN: My wife
took that picture of me.

955
00:41:49,420 --> 00:41:52,660
But this is, as you
saw, a chapter that--

956
00:41:52,660 --> 00:41:54,900
it's actually going through
a far amount of revisions

957
00:41:54,900 --> 00:41:56,200
from the version you've got.

958
00:41:56,200 --> 00:42:01,330
But it's an attempt to kind
of wrestle with this movement

959
00:42:01,330 --> 00:42:02,950
here around the future of work.

960
00:42:05,620 --> 00:42:08,680
And this concept
that people are going

961
00:42:08,680 --> 00:42:10,720
to get displaced
by automation, this

962
00:42:10,720 --> 00:42:12,940
has been with us
for a long time.

963
00:42:12,940 --> 00:42:15,520
So the first major
episode is really

964
00:42:15,520 --> 00:42:22,250
in Britain in 1815 around
the Luddite movement,

965
00:42:22,250 --> 00:42:27,190
with weavers smashing automated
looms that are putting them out

966
00:42:27,190 --> 00:42:28,990
of work.

967
00:42:28,990 --> 00:42:33,370
And eventually, a whole
division of the British Army,

968
00:42:33,370 --> 00:42:36,280
like 15,000 British
soldiers, get

969
00:42:36,280 --> 00:42:41,110
pulled in to put down this
pretty significant movement

970
00:42:41,110 --> 00:42:42,130
in Britain.

971
00:42:42,130 --> 00:42:45,610
So that's the first episode.

972
00:42:45,610 --> 00:42:49,400
We regularly go through this
debate about every 30 years,

973
00:42:49,400 --> 00:42:49,900
roughly.

974
00:42:52,540 --> 00:42:55,400
In 1950, I mentioned
this before,

975
00:42:55,400 --> 00:42:57,490
but Norbert Wiener painted
this very dark vision

976
00:42:57,490 --> 00:43:00,160
of computers displacing people.

977
00:43:00,160 --> 00:43:03,850
It took us a long time to
evolve into Licklider's vision.

978
00:43:03,850 --> 00:43:06,340
In the 1960s, there
was huge anxiety

979
00:43:06,340 --> 00:43:09,160
about workforce automation, as
a number of new technologies

980
00:43:09,160 --> 00:43:11,620
are being introduced into
the industrial processes.

981
00:43:14,950 --> 00:43:17,820
In 2015, that concern
came right back,

982
00:43:17,820 --> 00:43:20,650
that a mix of artificial
intelligence, machine learning,

983
00:43:20,650 --> 00:43:24,610
robotics was going to
be very threatening.

984
00:43:24,610 --> 00:43:25,640
This is an old debate.

985
00:43:25,640 --> 00:43:29,860
So the famous economist John
Maynard Keynes once famously

986
00:43:29,860 --> 00:43:32,770
wrote, "Thus we have
been expressly evolved

987
00:43:32,770 --> 00:43:36,280
by nature with all our
impulses and deepest

988
00:43:36,280 --> 00:43:41,170
instincts for the purpose of
solving the economic problem.

989
00:43:41,170 --> 00:43:43,480
If the economic
problem is solved,

990
00:43:43,480 --> 00:43:47,660
making will be deprived of
its traditional purpose."

991
00:43:47,660 --> 00:43:49,400
In other words, our
lives are organized

992
00:43:49,400 --> 00:43:53,818
around our work in many,
many important kinds of ways.

993
00:43:53,818 --> 00:43:55,360
And that's how our
lives get meaning.

994
00:43:55,360 --> 00:44:01,150
And if we're blowing up the work
model, what's going to happen?

995
00:44:01,150 --> 00:44:03,370
But the important
point to realize here

996
00:44:03,370 --> 00:44:06,450
is, there's no sign of that yet.

997
00:44:06,450 --> 00:44:09,220
The American workweek is
47 hours at this point.

998
00:44:09,220 --> 00:44:12,820
So we're some distance
away from this.

999
00:44:12,820 --> 00:44:15,370
We have time to reflect
on this, if we ever indeed

1000
00:44:15,370 --> 00:44:19,750
come close to that kind
of level of realization.

1001
00:44:19,750 --> 00:44:22,120
I just wanted to expose you
to this famous Keynes point,

1002
00:44:22,120 --> 00:44:24,250
because Keynes is
thinking about this

1003
00:44:24,250 --> 00:44:26,350
in the midst of the Depression.

1004
00:44:26,350 --> 00:44:30,150
He's thinking about the
whole future of work.

1005
00:44:30,150 --> 00:44:33,440
So the backdrop that
we're facing now

1006
00:44:33,440 --> 00:44:35,540
is significant work disruption.

1007
00:44:35,540 --> 00:44:38,330
And we've talked a
lot about that today.

1008
00:44:38,330 --> 00:44:40,670
Half of the manufacturing
jobs, as we talked back

1009
00:44:40,670 --> 00:44:44,690
in class number three, were
lost between 2000 and 2010.

1010
00:44:44,690 --> 00:44:49,710
And we saw the data on
median income decline,

1011
00:44:49,710 --> 00:44:53,330
and we talked about
the barbell problem.

1012
00:44:53,330 --> 00:44:58,940
And Brynjolfsson and
McAfee paint a picture

1013
00:44:58,940 --> 00:45:02,330
of technological
job displacement

1014
00:45:02,330 --> 00:45:06,650
and describe the accelerating
round of IT technologies

1015
00:45:06,650 --> 00:45:09,170
that are headed
towards the workplace.

1016
00:45:09,170 --> 00:45:12,680
Now, when we read about
advanced manufacturing,

1017
00:45:12,680 --> 00:45:14,510
the fix for the
manufacturing sector

1018
00:45:14,510 --> 00:45:17,930
is full of new technologies
like 3D printing

1019
00:45:17,930 --> 00:45:23,180
and digital production and
advanced sensors and photonics,

1020
00:45:23,180 --> 00:45:29,390
and in advanced materials,
a raft of new technologies

1021
00:45:29,390 --> 00:45:31,520
that are also going to
be entering that sector

1022
00:45:31,520 --> 00:45:33,290
to make that significantly
more efficient

1023
00:45:33,290 --> 00:45:34,850
and get the US back in the game.

1024
00:45:34,850 --> 00:45:37,460
That's the whole object
of that exercise, right?

1025
00:45:37,460 --> 00:45:40,100
So there is a raft
of new technologies

1026
00:45:40,100 --> 00:45:43,340
that is headed from the
IT world and elsewhere

1027
00:45:43,340 --> 00:45:44,720
into the workplace
that we're all

1028
00:45:44,720 --> 00:45:47,990
going to have to reckon with.

1029
00:45:47,990 --> 00:45:50,060
And there are
technological dystopians.

1030
00:45:50,060 --> 00:45:55,700
So Martin Ford argues that there
is a system of "winner take all

1031
00:45:55,700 --> 00:45:59,390
distribution" that's evolving,
that the technology of software

1032
00:45:59,390 --> 00:46:03,530
towards monopoly and the ability
of computers to do more than

1033
00:46:03,530 --> 00:46:06,583
they're programmed for-- deep
learning, as it's called--

1034
00:46:06,583 --> 00:46:08,000
is going to push
out a whole lower

1035
00:46:08,000 --> 00:46:11,900
end, that this "winner
take all distribution"

1036
00:46:11,900 --> 00:46:14,390
is going to be very
disruptive in this society.

1037
00:46:14,390 --> 00:46:18,560
David Cowen argues
that the country

1038
00:46:18,560 --> 00:46:21,290
will be divided by this
technological advance

1039
00:46:21,290 --> 00:46:22,335
into two countries.

1040
00:46:22,335 --> 00:46:23,960
We'll have a developed
world, and we'll

1041
00:46:23,960 --> 00:46:27,120
have an undeveloped world
in the United States.

1042
00:46:27,120 --> 00:46:31,640
And indeed, Peter Temin
of MIT, the economist,

1043
00:46:31,640 --> 00:46:36,310
has just written a book
arguing that current inequality

1044
00:46:36,310 --> 00:46:38,810
in the United States
has reached such a level

1045
00:46:38,810 --> 00:46:40,550
that the right
comparison is to look

1046
00:46:40,550 --> 00:46:43,990
at developing world
economics in a US context

1047
00:46:43,990 --> 00:46:45,350
to understand it for you.

1048
00:46:45,350 --> 00:46:49,100
Just for example,
Germany, the lowest

1049
00:46:49,100 --> 00:46:56,090
20% compared to the
highest 20%, it's 1 to 4.

1050
00:46:56,090 --> 00:46:57,860
In the United States,
the lowest 20%

1051
00:46:57,860 --> 00:47:02,660
of the population income to
the highest 20% is now 8 to 1.

1052
00:47:02,660 --> 00:47:05,570
That's pretty dramatic
income inequality.

1053
00:47:05,570 --> 00:47:09,440
That's a pretty-- that's
different than how the US was

1054
00:47:09,440 --> 00:47:13,850
for the previous century.

1055
00:47:13,850 --> 00:47:19,300
In their Second
Machine Age book,

1056
00:47:19,300 --> 00:47:21,690
Erik and Andrew argue
that the low-skilled

1057
00:47:21,690 --> 00:47:23,580
replaced the
middle-skilled jobs that

1058
00:47:23,580 --> 00:47:26,220
were displaced by technology.

1059
00:47:26,220 --> 00:47:28,710
They argue for tax
policy as a fix here.

1060
00:47:28,710 --> 00:47:32,850
Although, as we've discussed
and Rasheed led us into it,

1061
00:47:32,850 --> 00:47:36,850
that politically is a
very tough proposition.

1062
00:47:36,850 --> 00:47:39,960
There are studies that
project very high levels

1063
00:47:39,960 --> 00:47:41,860
of technological
job displacement.

1064
00:47:41,860 --> 00:47:45,450
So out of Oxford,
Frey and Osborne

1065
00:47:45,450 --> 00:47:47,640
look at occupational
descriptions

1066
00:47:47,640 --> 00:47:51,840
and conclude, oh,
47% of all US jobs

1067
00:47:51,840 --> 00:47:56,790
have high likelihood of
being replaced by automation.

1068
00:47:56,790 --> 00:48:00,870
Rob Atkinson at ITIF has
given a very strong critique

1069
00:48:00,870 --> 00:48:02,550
of that study.

1070
00:48:02,550 --> 00:48:08,280
He argues that it assumes
a highly unlikely 3%

1071
00:48:08,280 --> 00:48:11,887
labor productivity rate-- the
advent of all this automation.

1072
00:48:11,887 --> 00:48:13,720
We haven't seen that
since the 19th century.

1073
00:48:13,720 --> 00:48:19,050
So don't expect it to happen
anytime soon, Atkinson argues.

1074
00:48:19,050 --> 00:48:22,950
They also argue that
Frey and Osborne engage

1075
00:48:22,950 --> 00:48:25,290
in a lump of labor fallacy.

1076
00:48:25,290 --> 00:48:30,630
In other words, they assume that
there's a fixed amount of work

1077
00:48:30,630 --> 00:48:33,300
and that the amount of
work doesn't grow spurred

1078
00:48:33,300 --> 00:48:34,440
by these new technologies.

1079
00:48:34,440 --> 00:48:36,905
That's what David Autor led
us to understand better,

1080
00:48:36,905 --> 00:48:38,030
this whole complementarity.

1081
00:48:38,030 --> 00:48:41,430
In other words, there isn't
a fixed amount of work.

1082
00:48:41,430 --> 00:48:44,670
The complementarity
can create more,

1083
00:48:44,670 --> 00:48:47,520
so that there are larger
net potential gains here

1084
00:48:47,520 --> 00:48:50,290
across the economy.

1085
00:48:50,290 --> 00:48:53,350
The most realistic study of
technological displacement

1086
00:48:53,350 --> 00:48:57,480
so far has been by the
OECD done last summer.

1087
00:48:57,480 --> 00:49:01,200
They looked at 22 OECD nations.

1088
00:49:01,200 --> 00:49:06,240
And rather than look at
occupational descriptions,

1089
00:49:06,240 --> 00:49:10,440
they decided, let's go talk to
the workers that are actually

1090
00:49:10,440 --> 00:49:11,580
doing those jobs.

1091
00:49:11,580 --> 00:49:16,230
And this went through 22
different OECD countries.

1092
00:49:16,230 --> 00:49:18,690
They actually went out and
talked to the workforce.

1093
00:49:18,690 --> 00:49:20,600
And they would talk
to people about,

1094
00:49:20,600 --> 00:49:22,210
you may have a
particular job title,

1095
00:49:22,210 --> 00:49:24,030
but what are you actually doing?

1096
00:49:24,030 --> 00:49:25,890
What does your work
actually consistent of?

1097
00:49:25,890 --> 00:49:29,850
And of course, it turns out, and
we all surmise this in a way,

1098
00:49:29,850 --> 00:49:32,250
people are doing many things
that aren't necessarily

1099
00:49:32,250 --> 00:49:34,170
captioned in their job title.

1100
00:49:34,170 --> 00:49:37,740
Heaven forbid if somebody
held me to my job title.

1101
00:49:37,740 --> 00:49:39,730
Nothing would ever get done.

1102
00:49:39,730 --> 00:49:41,940
And we all know this, right?

1103
00:49:41,940 --> 00:49:44,140
And that's what the
OECD found-- people

1104
00:49:44,140 --> 00:49:47,010
are doing much more stuff than
their particularly narrow job

1105
00:49:47,010 --> 00:49:49,440
description may say
that they're doing.

1106
00:49:49,440 --> 00:49:52,710
So they concluded over an
extended period of time,

1107
00:49:52,710 --> 00:49:55,680
across those
developed countries,

1108
00:49:55,680 --> 00:49:58,830
the effect of technological
displacement was maybe 9%.

1109
00:49:58,830 --> 00:50:00,420
Now, that's a
significant number,

1110
00:50:00,420 --> 00:50:03,120
even over an extended
period of time.

1111
00:50:03,120 --> 00:50:04,800
That's a significant number.

1112
00:50:04,800 --> 00:50:08,520
But it's not 47%.

1113
00:50:08,520 --> 00:50:10,230
In the United States,
the number was 10%.

1114
00:50:10,230 --> 00:50:12,245
The range was 6%
to 12%, depending

1115
00:50:12,245 --> 00:50:14,370
on what kind of employment
you had in your country.

1116
00:50:17,310 --> 00:50:20,010
That means we have
a job ahead of us,

1117
00:50:20,010 --> 00:50:22,530
but it's not a nightmare.

1118
00:50:22,530 --> 00:50:26,070
It's not something we should
be sleepless over every night

1119
00:50:26,070 --> 00:50:28,170
from now on.

1120
00:50:28,170 --> 00:50:31,980
There's a recent effort on--
this is not in the book,

1121
00:50:31,980 --> 00:50:32,700
in the chapter--

1122
00:50:32,700 --> 00:50:36,390
but Daron Acemoglu, another
very noted MIT economist

1123
00:50:36,390 --> 00:50:39,180
who does a lot of
work with David Autor

1124
00:50:39,180 --> 00:50:41,370
and is a real student
of innovation,

1125
00:50:41,370 --> 00:50:46,680
he's a growth economist,
Daron looked at robotics.

1126
00:50:46,680 --> 00:50:48,900
And he looked at
job displacement

1127
00:50:48,900 --> 00:50:49,920
in the robotics sector.

1128
00:50:49,920 --> 00:50:51,930
And he had a pretty useful--

1129
00:50:51,930 --> 00:50:54,390
had a good database that
he was working from.

1130
00:50:54,390 --> 00:51:02,500
He concluded that over a
17-year period ending in 1977,

1131
00:51:02,500 --> 00:51:04,510
a total for the
entire United States

1132
00:51:04,510 --> 00:51:08,680
economy of technological job
displacement caused by robotics

1133
00:51:08,680 --> 00:51:12,100
was somewhere between
300,000 and 600,000 jobs

1134
00:51:12,100 --> 00:51:14,020
over a 17-year period.

1135
00:51:14,020 --> 00:51:16,600
That is not a big number.

1136
00:51:16,600 --> 00:51:18,580
Job churn in the
United States per week

1137
00:51:18,580 --> 00:51:23,290
is something like
75,000 jobs per week.

1138
00:51:23,290 --> 00:51:25,190
Look at it in that
kind of context.

1139
00:51:25,190 --> 00:51:29,170
So these changes are
going to be significant.

1140
00:51:29,170 --> 00:51:32,320
And I would argue
they will occur.

1141
00:51:32,320 --> 00:51:36,220
But we may have a time period
that we're able to look at them

1142
00:51:36,220 --> 00:51:40,240
and think and plan and act
rather than being completely

1143
00:51:40,240 --> 00:51:42,430
disruptive overnight.

1144
00:51:42,430 --> 00:51:48,280
So there's a quote
by the vice president

1145
00:51:48,280 --> 00:51:52,960
of the Danish Conference
of Trade Unions named

1146
00:51:52,960 --> 00:51:55,780
Nauna Hejlund.

1147
00:51:55,780 --> 00:51:59,080
Her comment at an
OECD conference

1148
00:51:59,080 --> 00:52:03,550
I went to last fall
was, "New technology

1149
00:52:03,550 --> 00:52:07,270
is not the enemy of
workers, old technology is."

1150
00:52:07,270 --> 00:52:11,500
And what she's saying here
is that the companies most

1151
00:52:11,500 --> 00:52:14,950
likely to fail, where
all jobs will be lost,

1152
00:52:14,950 --> 00:52:18,520
are those who are not keeping
up with the technology.

1153
00:52:18,520 --> 00:52:22,150
That workers have a stake in
keeping their firms current

1154
00:52:22,150 --> 00:52:23,380
with current realities.

1155
00:52:29,970 --> 00:52:32,440
And here's a chart
that just shows

1156
00:52:32,440 --> 00:52:34,000
the rate at which
robots are being

1157
00:52:34,000 --> 00:52:35,500
installed into production.

1158
00:52:35,500 --> 00:52:38,680
It's actually slowed
fairly significantly.

1159
00:52:38,680 --> 00:52:40,390
Now, look, we're in
a process of moving

1160
00:52:40,390 --> 00:52:43,690
from old-style,
industrial robotics, which

1161
00:52:43,690 --> 00:52:48,210
used to weigh tons and
they were fixed in place

1162
00:52:48,210 --> 00:52:49,960
and they had to be
fenced off, because you

1163
00:52:49,960 --> 00:52:52,480
didn't want to be anywhere near
them, because they were unsafe.

1164
00:52:52,480 --> 00:52:53,938
And they would do
the perfect weld,

1165
00:52:53,938 --> 00:52:56,967
and that's all they
could do, on an oil line.

1166
00:52:56,967 --> 00:52:59,050
We're moving to a whole
new generation of robotics

1167
00:52:59,050 --> 00:53:00,700
that are much more
flexible, that

1168
00:53:00,700 --> 00:53:04,770
are subject to
voice command, that

1169
00:53:04,770 --> 00:53:06,625
are a radically different
kind of robotics

1170
00:53:06,625 --> 00:53:08,000
than the old
industrial robotics.

1171
00:53:08,000 --> 00:53:12,070
So some of this data
may reflect the fact

1172
00:53:12,070 --> 00:53:14,080
that that new
generation of robotics

1173
00:53:14,080 --> 00:53:17,350
is only now just starting to be
thought about and put together.

1174
00:53:17,350 --> 00:53:17,850
Max?

1175
00:53:17,850 --> 00:53:19,392
MAX: Do you think
that this has to do

1176
00:53:19,392 --> 00:53:21,190
with some of the
anxieties that workers

1177
00:53:21,190 --> 00:53:23,620
have expressed around
being replaced by robotics?

1178
00:53:23,620 --> 00:53:26,185
Like there might be some sort
of correlation, relation there?

1179
00:53:26,185 --> 00:53:27,810
WILLIAM BONVILLIAN:
This slowdown rate?

1180
00:53:27,810 --> 00:53:28,485
MAX: Yeah.

1181
00:53:28,485 --> 00:53:29,860
WILLIAM BONVILLIAN:
I don't know.

1182
00:53:29,860 --> 00:53:32,410
I don't think so,
because we don't have

1183
00:53:32,410 --> 00:53:34,360
an organized workforce anymore.

1184
00:53:34,360 --> 00:53:35,860
We're down to about
11% unionization

1185
00:53:35,860 --> 00:53:37,630
in the United States.

1186
00:53:37,630 --> 00:53:38,170
MAX: Really?

1187
00:53:38,170 --> 00:53:39,170
WILLIAM BONVILLIAN: Yes.

1188
00:53:39,170 --> 00:53:43,390
So workers don't have a lot
to say at this point in the US

1189
00:53:43,390 --> 00:53:44,638
about a lot of these changes.

1190
00:53:44,638 --> 00:53:46,930
They don't have a lot of
control over their employment,

1191
00:53:46,930 --> 00:53:52,550
frankly, which is
another set of issues.

1192
00:53:52,550 --> 00:53:58,840
But the chapter argues
that the most immediate

1193
00:53:58,840 --> 00:54:04,900
near-term problem is not
technological displacement.

1194
00:54:04,900 --> 00:54:07,000
That's going to be a
longer-term problem.

1195
00:54:07,000 --> 00:54:10,600
The most immediate
near-term problem

1196
00:54:10,600 --> 00:54:14,440
is what economists call
"secular stagnation."

1197
00:54:14,440 --> 00:54:16,240
And that's the
problem we're in now.

1198
00:54:16,240 --> 00:54:21,580
And this term was developed
by Larry Summers in 2013

1199
00:54:21,580 --> 00:54:24,100
to essentially
describe this situation

1200
00:54:24,100 --> 00:54:26,680
that we've got at the moment--

1201
00:54:26,680 --> 00:54:29,260
interest rates around
zero and US output

1202
00:54:29,260 --> 00:54:31,600
is insufficient to
support full employment.

1203
00:54:31,600 --> 00:54:32,600
He's running in 2013.

1204
00:54:32,600 --> 00:54:33,850
It's gotten better since then.

1205
00:54:33,850 --> 00:54:36,370
But we do have this
structural unemployment,

1206
00:54:36,370 --> 00:54:40,330
where a significant number of
people who had jobs no longer

1207
00:54:40,330 --> 00:54:41,590
are in the workforce.

1208
00:54:41,590 --> 00:54:44,440
That's the problem
that's still with us.

1209
00:54:44,440 --> 00:54:46,510
He argues that there
are a series of factors

1210
00:54:46,510 --> 00:54:49,030
that affect this, that are
hampering the investment

1211
00:54:49,030 --> 00:54:50,020
demand.

1212
00:54:50,020 --> 00:54:52,630
So decreasing
population growth--

1213
00:54:52,630 --> 00:54:55,660
we can all see that-- a relative
decrease in the cost of capital

1214
00:54:55,660 --> 00:54:58,450
goods, excess money being
retained and not used

1215
00:54:58,450 --> 00:55:00,610
by large corporations--
so retained earnings

1216
00:55:00,610 --> 00:55:03,370
by large companies is
very high at this point.

1217
00:55:03,370 --> 00:55:06,520
On the savings side,
there's, he would argue,

1218
00:55:06,520 --> 00:55:10,990
excessive reserved holdings
in developing countries

1219
00:55:10,990 --> 00:55:15,550
due to, in part, post-crisis
financial regulation,

1220
00:55:15,550 --> 00:55:20,200
inequality is a cause here, and
increasing intermediation costs

1221
00:55:20,200 --> 00:55:24,520
are all causative factors in
this period of low growth.

1222
00:55:24,520 --> 00:55:28,780
We have very low GDP, and we
have very low productivity.

1223
00:55:28,780 --> 00:55:32,770
So economy-wide, our
productivity is about 1.12%.

1224
00:55:32,770 --> 00:55:34,265
Not a good level, right?

1225
00:55:34,265 --> 00:55:36,340
MAX: But we still
have low unemployment.

1226
00:55:36,340 --> 00:55:36,590
WILLIAM BONVILLIAN: Right?

1227
00:55:36,590 --> 00:55:38,200
MAX: But we still
have low unemployment.

1228
00:55:38,200 --> 00:55:40,370
WILLIAM BONVILLIAN: Yeah,
we have 4.9% unemployment.

1229
00:55:40,370 --> 00:55:43,510
But go back to Rasheed's
barbell example--

1230
00:55:43,510 --> 00:55:44,980
a lot of that is
the thinning out

1231
00:55:44,980 --> 00:55:48,440
of the center being pushed
to lower-end services jobs.

1232
00:55:48,440 --> 00:55:51,310
RASHEED: Also, the labor force
participation rate is too high.

1233
00:55:51,310 --> 00:55:52,390
WILLIAM BONVILLIAN: Yeah.

1234
00:55:52,390 --> 00:55:56,710
So we have low productivity,
low capital investment,

1235
00:55:56,710 --> 00:56:00,160
and a low growth rate, growing
inequality, and a declining

1236
00:56:00,160 --> 00:56:01,570
middle class.

1237
00:56:01,570 --> 00:56:04,040
These are big problems.

1238
00:56:04,040 --> 00:56:07,930
Now, low productivity rate,
low capital investment,

1239
00:56:07,930 --> 00:56:13,360
that tells us that automation
is not happening at the moment,

1240
00:56:13,360 --> 00:56:16,840
because automation is designed
to increase productivity

1241
00:56:16,840 --> 00:56:19,060
and is based on increased
capital investment.

1242
00:56:19,060 --> 00:56:20,750
We're not seeing
either one of those,

1243
00:56:20,750 --> 00:56:23,680
so that tells us that
we're not entering

1244
00:56:23,680 --> 00:56:27,030
a period of radical
growth in automation.

1245
00:56:27,030 --> 00:56:27,530
Martin?

1246
00:56:27,530 --> 00:56:29,800
MARTIN: So is this just
the US or is it global?

1247
00:56:29,800 --> 00:56:32,590
And also, is it taking into
account the industrialization

1248
00:56:32,590 --> 00:56:34,096
jobs leaving?

1249
00:56:34,096 --> 00:56:35,830
WILLIAM BONVILLIAN: Yeah.

1250
00:56:35,830 --> 00:56:37,540
It does take into
account the decline

1251
00:56:37,540 --> 00:56:39,650
in manufacturing employment.

1252
00:56:39,650 --> 00:56:41,830
This is a US picture
that Summers is painting,

1253
00:56:41,830 --> 00:56:44,500
and he's going to propose a US
solution in a second, which is

1254
00:56:44,500 --> 00:56:46,720
a major infrastructure package.

1255
00:56:46,720 --> 00:56:51,910
But the secular stagnation
is a phenomena now

1256
00:56:51,910 --> 00:56:55,300
in European economies as well,
and in the Japanese economy.

1257
00:56:55,300 --> 00:56:58,420
So this is a worldwide
phenomena of the developed

1258
00:56:58,420 --> 00:57:00,040
world at the moment.

1259
00:57:00,040 --> 00:57:02,680
And it's a problem
that you all have.

1260
00:57:02,680 --> 00:57:05,440
You're going to have this
future work problem for sure,

1261
00:57:05,440 --> 00:57:06,890
but now you've got this one.

1262
00:57:06,890 --> 00:57:09,400
AUDIENCE: Do we also
have a aging workforce?

1263
00:57:09,400 --> 00:57:11,560
WILLIAM BONVILLIAN: Yes,
aging workforce, right.

1264
00:57:11,560 --> 00:57:14,030
And that's a powerful
part of this story, too.

1265
00:57:14,030 --> 00:57:14,530
OK?

1266
00:57:14,530 --> 00:57:23,160
So if you have a population
increase, a 1% year population

1267
00:57:23,160 --> 00:57:27,120
increase, that guarantees
you one additional percentage

1268
00:57:27,120 --> 00:57:29,130
point in your growth rate.

1269
00:57:29,130 --> 00:57:30,810
If you have zero
population increase,

1270
00:57:30,810 --> 00:57:34,860
you just pull one percentage of
growth out of your growth rate,

1271
00:57:34,860 --> 00:57:36,450
out of your GDP rate.

1272
00:57:36,450 --> 00:57:39,200
So that's a phenomena that a lot
of these developing countries

1273
00:57:39,200 --> 00:57:40,590
are starting to wrestle with.

1274
00:57:40,590 --> 00:57:42,673
And China's going to have
to wrestle with this one

1275
00:57:42,673 --> 00:57:44,730
fairly soon.

1276
00:57:44,730 --> 00:57:47,580
So here's what
Summers is talking

1277
00:57:47,580 --> 00:57:51,180
about in terms of excessive
savings over investment.

1278
00:57:51,180 --> 00:57:53,910
Look at the split between
savings and investment levels.

1279
00:57:53,910 --> 00:57:58,140
That means we're not investing
our savings in the society,

1280
00:57:58,140 --> 00:57:59,640
in the economy.

1281
00:57:59,640 --> 00:58:01,380
And Summers identifies this gap.

1282
00:58:01,380 --> 00:58:06,680
RASHEED: And in 2008
it's like where?

1283
00:58:06,680 --> 00:58:08,350
WILLIAM BONVILLIAN:
Right there, right?

1284
00:58:08,350 --> 00:58:10,960
Well, that's when
savings falls apart.

1285
00:58:10,960 --> 00:58:14,140
But there's still a
big gap in the system.

1286
00:58:14,140 --> 00:58:17,170
And investment hits then, too.

1287
00:58:17,170 --> 00:58:19,630
I want you to know of this
economist named Robert Gordon,

1288
00:58:19,630 --> 00:58:22,210
because you all are
hanging out at MIT,

1289
00:58:22,210 --> 00:58:25,930
and we believe that
technological advance is

1290
00:58:25,930 --> 00:58:28,360
the most powerful thing ever.

1291
00:58:28,360 --> 00:58:31,420
But Robert Gordon paints
a different picture.

1292
00:58:31,420 --> 00:58:33,460
And you just ought to be
aware of this debate--

1293
00:58:33,460 --> 00:58:35,500
not that he's necessarily right.

1294
00:58:35,500 --> 00:58:39,310
This was by far the
best-selling book on economics

1295
00:58:39,310 --> 00:58:41,170
for the last two years.

1296
00:58:41,170 --> 00:58:43,840
And Gordon's book is
called The Rise and Fall

1297
00:58:43,840 --> 00:58:44,770
of American Growth.

1298
00:58:44,770 --> 00:58:47,875
He argues that the current
low-growth secular stagnation,

1299
00:58:47,875 --> 00:58:51,280
and he uses the
term, is the result

1300
00:58:51,280 --> 00:58:56,680
not of insufficient demand, but
of insufficient supply, i.e.

1301
00:58:56,680 --> 00:58:59,380
insufficient
technological supply.

1302
00:58:59,380 --> 00:59:04,330
So he argues that the
IT revolution is fading

1303
00:59:04,330 --> 00:59:07,120
and that we never got as
much out of the IT revolution

1304
00:59:07,120 --> 00:59:11,140
as we did out of prior
innovation waves.

1305
00:59:11,140 --> 00:59:13,870
This one just wasn't that
big compared to electricity

1306
00:59:13,870 --> 00:59:16,480
and compared to railroads and
some of the other big waves

1307
00:59:16,480 --> 00:59:18,250
that we've been through.

1308
00:59:18,250 --> 00:59:22,930
And that we're locked on
to an innovation wave that

1309
00:59:22,930 --> 00:59:25,030
is just not, at this
stage, producing

1310
00:59:25,030 --> 00:59:28,060
that much productivity growth.

1311
00:59:28,060 --> 00:59:31,990
And he goes into enormous
depth about each one

1312
00:59:31,990 --> 00:59:32,930
of these technologies.

1313
00:59:32,930 --> 00:59:36,010
So a complete rarity
for an economist-- he

1314
00:59:36,010 --> 00:59:39,850
actually has read into
the technology literature

1315
00:59:39,850 --> 00:59:42,190
and is looking at what
these intentions actually

1316
00:59:42,190 --> 00:59:44,660
were in the second half
of the 19th century

1317
00:59:44,660 --> 00:59:46,370
and what they meant.

1318
00:59:46,370 --> 00:59:50,748
So it's an intriguing
alternative story,

1319
00:59:50,748 --> 00:59:52,540
one that's contrary to
the story that we've

1320
00:59:52,540 --> 00:59:55,280
been telling ourselves
for a long period of time.

1321
00:59:55,280 --> 00:59:56,410
How do we fix this?

1322
00:59:59,440 --> 01:00:03,760
Gordon says that it involves--
fixing societal headwinds is

1323
01:00:03,760 --> 01:00:05,920
the best way out of this box.

1324
01:00:05,920 --> 01:00:10,830
And improving growth will depend
on educational attainment.

1325
01:00:10,830 --> 01:00:12,580
And remember when we
read Goldin and Katz,

1326
01:00:12,580 --> 01:00:17,080
we talked about how a key to the
historic American growth rate

1327
01:00:17,080 --> 01:00:20,920
was that there's always
a rise in increased

1328
01:00:20,920 --> 01:00:23,507
technical requirements
in the economy?

1329
01:00:23,507 --> 01:00:25,090
And the genius of
the American system,

1330
01:00:25,090 --> 01:00:29,140
at least till the mid-'70s, was
that we kept an education curve

1331
01:00:29,140 --> 01:00:31,630
between high school, and
particularly Mass college

1332
01:00:31,630 --> 01:00:34,660
education, out ahead
of that, right?

1333
01:00:34,660 --> 01:00:37,720
That, in turn-- those curves
are related to each other.

1334
01:00:37,720 --> 01:00:40,480
That talent base, as
we know from Romer,

1335
01:00:40,480 --> 01:00:42,310
helps drive the
technological base.

1336
01:00:42,310 --> 01:00:44,130
If you get a large
part of your economy

1337
01:00:44,130 --> 01:00:48,040
out in prospecting or involved
in supporting a technology

1338
01:00:48,040 --> 01:00:50,830
advance, the whole
system will do better.

1339
01:00:50,830 --> 01:00:55,690
So when we leveled out the
graduation rate at the college

1340
01:00:55,690 --> 01:01:00,790
level in the mid-'70s, we paid
a big price in terms of ongoing

1341
01:01:00,790 --> 01:01:03,845
technology productivity and
technology advance, as well.

1342
01:01:03,845 --> 01:01:05,470
So that's kind of
what he's suggesting.

1343
01:01:05,470 --> 01:01:08,290
It's an interesting-- he doesn't
spell it out quite that way.

1344
01:01:08,290 --> 01:01:10,850
But it's an interesting idea.

1345
01:01:10,850 --> 01:01:13,900
So tackling some of these
underlying social problems,

1346
01:01:13,900 --> 01:01:16,390
including inequality,
i.e. bring more people

1347
01:01:16,390 --> 01:01:22,090
into the economy, all of
these are fixes, in his mind,

1348
01:01:22,090 --> 01:01:27,100
for getting around the fact
that the IT revolution isn't

1349
01:01:27,100 --> 01:01:28,930
quite what we hoped for.

1350
01:01:28,930 --> 01:01:32,540
Now, what do I think of that?

1351
01:01:32,540 --> 01:01:34,240
You know, time will tell.

1352
01:01:34,240 --> 01:01:35,300
We're going to find out.

1353
01:01:35,300 --> 01:01:38,110
I think it may be we had
the rapid rise, we're

1354
01:01:38,110 --> 01:01:40,630
on a more moderate growth
pattern in the IT revolution,

1355
01:01:40,630 --> 01:01:43,240
as we talked about we talked
about innovation waves.

1356
01:01:43,240 --> 01:01:45,160
And we can only expect
more moderate growth

1357
01:01:45,160 --> 01:01:46,840
as a result of that.

1358
01:01:46,840 --> 01:01:49,120
But we are waiting for
the next big thing,

1359
01:01:49,120 --> 01:01:50,778
for the next big innovation.

1360
01:01:53,710 --> 01:01:58,390
So Summers' solution is to
throw money in infrastructure--

1361
01:01:58,390 --> 01:02:00,010
and, well, let's do roads--

1362
01:02:00,010 --> 01:02:01,850
and that will help
that community

1363
01:02:01,850 --> 01:02:04,030
that got displaced in
manufacturing jobs that

1364
01:02:04,030 --> 01:02:06,490
aren't upskilling.

1365
01:02:06,490 --> 01:02:08,120
That could help them.

1366
01:02:08,120 --> 01:02:11,880
But the problem with Summers'
solution, I would argue

1367
01:02:11,880 --> 01:02:14,740
and the chapter argues, is
that it doesn't do anything

1368
01:02:14,740 --> 01:02:18,670
to address productivity, which
is the heart of the problem.

1369
01:02:18,670 --> 01:02:25,690
Getting that productivity number
up, as we know, increases GDP.

1370
01:02:25,690 --> 01:02:28,030
It's a key causative factor
behind the growth rate.

1371
01:02:28,030 --> 01:02:30,100
And let's get that
growth rate up

1372
01:02:30,100 --> 01:02:32,590
and overall economic
well-being will improve,

1373
01:02:32,590 --> 01:02:34,590
and you'll have more money
to distribute, right?

1374
01:02:34,590 --> 01:02:36,730
Let's at least get that right.

1375
01:02:36,730 --> 01:02:38,980
But an infrastructure
program that's

1376
01:02:38,980 --> 01:02:43,390
aimed at roads and sewers, we
already have our road system.

1377
01:02:43,390 --> 01:02:45,730
We have 90,000 miles
of interstate highway.

1378
01:02:45,730 --> 01:02:48,700
We're not going to
get, by repairing it,

1379
01:02:48,700 --> 01:02:52,000
we're not going to get
the innovation wave boost

1380
01:02:52,000 --> 01:02:54,490
that the combination of
building that interstate system

1381
01:02:54,490 --> 01:02:57,160
and internal combustion
engines got us back

1382
01:02:57,160 --> 01:02:58,780
in the '40s and '50s.

1383
01:02:58,780 --> 01:03:00,700
We're just not going
to get that out of it.

1384
01:03:00,700 --> 01:03:04,260
So Summers' prescription of
just doing infrastructure

1385
01:03:04,260 --> 01:03:08,980
will help people who
are underemployed,

1386
01:03:08,980 --> 01:03:12,160
but it's not going to
wrestle with this underlying

1387
01:03:12,160 --> 01:03:16,930
productivity rate, which is
really the problem at hand.

1388
01:03:16,930 --> 01:03:19,540
And the alternative
here, the book argues,

1389
01:03:19,540 --> 01:03:21,505
is investing in
innovation infrastructure.

1390
01:03:21,505 --> 01:03:23,740
In other words,
having a larger view

1391
01:03:23,740 --> 01:03:25,710
of what that
infrastructure ought to be

1392
01:03:25,710 --> 01:03:28,780
and investing in that.

1393
01:03:28,780 --> 01:03:29,680
All right.

1394
01:03:29,680 --> 01:03:32,010
That's this text.

1395
01:03:32,010 --> 01:03:33,940
And let's take it apart.

1396
01:03:33,940 --> 01:03:35,360
Luyao, it's all yours.

1397
01:03:38,440 --> 01:03:40,890
I'll go back to the hiking
picture at the beginning.

1398
01:03:40,890 --> 01:03:42,140
RASHEED: Everybody's favorite.

1399
01:03:42,140 --> 01:03:44,370
MARTIN: So it's the last
16 minutes, pretty much.

1400
01:03:44,370 --> 01:03:47,852
So do you want to wrap up, and
then go back to the discussion?

1401
01:03:47,852 --> 01:03:49,310
WILLIAM BONVILLIAN:
Well, let's let

1402
01:03:49,310 --> 01:03:51,440
Luyao tell us some
key observations

1403
01:03:51,440 --> 01:03:52,480
about this chapter.

1404
01:03:52,480 --> 01:03:55,450
I agree, we're heading
towards the deadline,

1405
01:03:55,450 --> 01:03:57,880
and there's some things
I wanted to say, too.

1406
01:03:57,880 --> 01:04:00,020
But go ahead.

1407
01:04:00,020 --> 01:04:02,150
AUDIENCE: Yeah, we
realize there's basically

1408
01:04:02,150 --> 01:04:04,130
the competition
of the technology

1409
01:04:04,130 --> 01:04:06,590
replacement and the
possible solutions

1410
01:04:06,590 --> 01:04:10,460
that we have, and also all
the policy factors that we

1411
01:04:10,460 --> 01:04:12,380
want to take into account.

1412
01:04:12,380 --> 01:04:15,950
And I think the technology
replacement can actually

1413
01:04:15,950 --> 01:04:21,320
be an opportunity for economies
having an aging population.

1414
01:04:21,320 --> 01:04:23,210
But at the same time,
the United States

1415
01:04:23,210 --> 01:04:29,400
facing low investment
rate and low savings

1416
01:04:29,400 --> 01:04:33,240
and low economic
growth is probably

1417
01:04:33,240 --> 01:04:38,670
problematic for this
substitution to work well.

1418
01:04:42,660 --> 01:04:53,137
So what are the suggestions
that we have to just improve--

1419
01:04:53,137 --> 01:04:54,220
maybe this is too general.

1420
01:05:00,260 --> 01:05:03,243
Yeah, so what are the--

1421
01:05:03,243 --> 01:05:04,410
well, let's talk about this.

1422
01:05:04,410 --> 01:05:07,680
So in one of the
suggestions it's

1423
01:05:07,680 --> 01:05:12,130
saying by reducing the
number of working hours,

1424
01:05:12,130 --> 01:05:16,710
it can actually help to
encourage consumption and help

1425
01:05:16,710 --> 01:05:17,940
to solve unemployment.

1426
01:05:17,940 --> 01:05:20,000
And how do you guys
like this idea?

1427
01:05:20,000 --> 01:05:21,870
MAX: Of reducing hours?

1428
01:05:21,870 --> 01:05:24,840
AUDIENCE: Yeah, reducing
the number of working hours.

1429
01:05:24,840 --> 01:05:25,530
WILLIAM BONVILLIAN:
France has been

1430
01:05:25,530 --> 01:05:26,610
working on this for a while.

1431
01:05:26,610 --> 01:05:27,690
LILY: I would love that.

1432
01:05:27,690 --> 01:05:30,510
But then I think you always
have the people who--

1433
01:05:30,510 --> 01:05:32,940
I mean, how do you enforce it
and how do you regulate it?

1434
01:05:32,940 --> 01:05:35,580
Because you'll always
have the people who will

1435
01:05:35,580 --> 01:05:37,857
want to work more
hours to get ahead.

1436
01:05:37,857 --> 01:05:39,190
That's one of the reasons that--

1437
01:05:39,190 --> 01:05:39,330
AUDIENCE: Yeah.

1438
01:05:39,330 --> 01:05:40,663
That's what Chinese do, I guess.

1439
01:05:40,663 --> 01:05:41,440
[LAUGHTER]

1440
01:05:41,440 --> 01:05:43,320
MAX: Yeah, you
define when people

1441
01:05:43,320 --> 01:05:44,508
start getting paid overtime.

1442
01:05:44,508 --> 01:05:45,300
It's not that hard.

1443
01:05:45,300 --> 01:05:47,310
And then, make it
so that overtime,

1444
01:05:47,310 --> 01:05:51,480
as it currently is, you
have to get permission

1445
01:05:51,480 --> 01:05:53,950
from your supervisor in order
to actually get those hours.

1446
01:05:53,950 --> 01:05:55,680
AUDIENCE: At most companies,
that's already in place.

1447
01:05:55,680 --> 01:05:58,010
If you're going overtime,
you need an excuse for it.

1448
01:05:58,010 --> 01:05:58,270
MAX: Exactly.

1449
01:05:58,270 --> 01:05:59,630
AUDIENCE: And people just do
it under the table instead.

1450
01:05:59,630 --> 01:06:00,900
Because by billing
overtime, you're

1451
01:06:00,900 --> 01:06:03,400
admitting you can't finish your
work in the normal hours, so

1452
01:06:03,400 --> 01:06:04,190
[INAUDIBLE].

1453
01:06:04,190 --> 01:06:05,730
LILY: Yeah.

1454
01:06:05,730 --> 01:06:09,360
So in my field, we're
not paid hourly.

1455
01:06:09,360 --> 01:06:11,790
We're paid on--
the way that we are

1456
01:06:11,790 --> 01:06:16,650
measured is in the output of
publications that we have.

1457
01:06:16,650 --> 01:06:18,780
So, yeah, I didn't want
to work until 1 o'clock

1458
01:06:18,780 --> 01:06:21,330
in the morning three
nights last week.

1459
01:06:21,330 --> 01:06:23,730
But I did, because I had
a paper I had to get out.

1460
01:06:23,730 --> 01:06:27,600
And so I can't
imagine, in my field,

1461
01:06:27,600 --> 01:06:30,533
an hour limit being in place.

1462
01:06:30,533 --> 01:06:32,700
MAX: Well, yeah, but that's
different from a factory

1463
01:06:32,700 --> 01:06:35,893
job or some sort of--

1464
01:06:35,893 --> 01:06:38,060
MARTIN: Yeah, I think
cultural context is important.

1465
01:06:38,060 --> 01:06:41,350
Just like I don't
think it's very--

1466
01:06:41,350 --> 01:06:43,480
I don't think I went
into the USC to do less.

1467
01:06:43,480 --> 01:06:43,760
WILLIAM BONVILLIAN: Right.

1468
01:06:43,760 --> 01:06:46,170
This is not a very American
solution, I would say.

1469
01:06:46,170 --> 01:06:46,990
[LAUGHTER]

1470
01:06:46,990 --> 01:06:48,260
MARTIN: I would try
to think about-- well,

1471
01:06:48,260 --> 01:06:48,950
maybe there's certain sectors--

1472
01:06:48,950 --> 01:06:50,210
WILLIAM BONVILLIAN: Because
we are driven workaholics.

1473
01:06:50,210 --> 01:06:51,520
MARTIN: Yeah.

1474
01:06:51,520 --> 01:06:53,772
It is very much the culture.

1475
01:06:53,772 --> 01:06:55,520
STEPH: Well, in
the future of work,

1476
01:06:55,520 --> 01:06:58,520
one of the trends I've seen
is people talking about who

1477
01:06:58,520 --> 01:07:00,110
is going to lose jobs.

1478
01:07:00,110 --> 01:07:02,270
And they always say
that the alternative

1479
01:07:02,270 --> 01:07:04,820
is that the people who are
going to gain jobs are creatives

1480
01:07:04,820 --> 01:07:06,770
and that it's the
creative economy that's

1481
01:07:06,770 --> 01:07:10,538
going to gain the most from
automation in manufacturing.

1482
01:07:10,538 --> 01:07:12,830
So there's going to be really
interesting opportunities

1483
01:07:12,830 --> 01:07:14,690
in, say, virtual
reality, for people

1484
01:07:14,690 --> 01:07:16,820
who are interested in
producing, say, entertainment

1485
01:07:16,820 --> 01:07:21,140
specifically for that
technological environment.

1486
01:07:21,140 --> 01:07:24,420
AUDIENCE: I think Romer proposed
that there will be more job

1487
01:07:24,420 --> 01:07:25,670
placement in the middle class.

1488
01:07:25,670 --> 01:07:30,590
And we'll see a higher demand
for high-pay, high-skilled jobs

1489
01:07:30,590 --> 01:07:33,540
and low-pay, low-skilled
jobs in this,

1490
01:07:33,540 --> 01:07:37,560
and we'll see a reduced
size of middle class.

1491
01:07:37,560 --> 01:07:42,970
So in terms of the rising demand
for the low-skill, low-pay

1492
01:07:42,970 --> 01:07:47,870
jobs, what would be the
social problems associated

1493
01:07:47,870 --> 01:07:51,220
with that kind of future?

1494
01:07:51,220 --> 01:07:54,860
RASHEED: Yeah, I think you just
highlighted a pretty important

1495
01:07:54,860 --> 01:07:57,830
point that we tried to get at,
which is just this hollowing

1496
01:07:57,830 --> 01:08:00,830
out of the middle class.

1497
01:08:00,830 --> 01:08:04,790
But more importantly,
the economic models

1498
01:08:04,790 --> 01:08:07,105
that we're going to have to
use are not going to be--

1499
01:08:07,105 --> 01:08:08,480
now, everyone's
income levels are

1500
01:08:08,480 --> 01:08:11,720
distributed relatively
evenly in accordance

1501
01:08:11,720 --> 01:08:12,830
to population numbers.

1502
01:08:12,830 --> 01:08:16,590
But you have to use models from
different countries and stuff

1503
01:08:16,590 --> 01:08:17,090
like that.

1504
01:08:17,090 --> 01:08:20,990
It's going to be like an
entirely different way to think

1505
01:08:20,990 --> 01:08:24,710
about the economy, because we've
never really had to deal with

1506
01:08:24,710 --> 01:08:28,473
an economy where we're
so-- now, we're like--

1507
01:08:28,473 --> 01:08:30,890
we've never been bimodal before,
having to deal with like,

1508
01:08:30,890 --> 01:08:32,689
we only have
low-skill, low-pay jobs

1509
01:08:32,689 --> 01:08:34,819
or high-skill, high-pay jobs.

1510
01:08:34,819 --> 01:08:37,668
And on the [INAUDIBLE]
the US sits on,

1511
01:08:37,668 --> 01:08:39,710
I don't think we've had
to deal with that before.

1512
01:08:39,710 --> 01:08:40,403
And so--

1513
01:08:40,403 --> 01:08:43,070
STEPH: I don't think a democracy
has had to deal with it before.

1514
01:08:43,070 --> 01:08:45,320
AUDIENCE: I mean, would you
not say that early America

1515
01:08:45,320 --> 01:08:46,290
was pretty bimodal?

1516
01:08:46,290 --> 01:08:48,140
You had plantation
owners, and you

1517
01:08:48,140 --> 01:08:50,270
had the bankers in New York.

1518
01:08:50,270 --> 01:08:51,748
Everyone else was pretty poor.

1519
01:08:51,748 --> 01:08:52,790
WILLIAM BONVILLIAN: Yeah.

1520
01:08:52,790 --> 01:08:54,710
I think there were
times in the later part

1521
01:08:54,710 --> 01:08:57,109
of the 19th century where we
had these kind of inequality

1522
01:08:57,109 --> 01:08:58,029
numbers, too.

1523
01:08:58,029 --> 01:08:59,720
AUDIENCE: Yeah, so I think
it was a huge problem--

1524
01:08:59,720 --> 01:09:00,710
WILLIAM BONVILLIAN:
But we got out of them.

1525
01:09:00,710 --> 01:09:03,590
We got out of them, essentially
by raising the education

1526
01:09:03,590 --> 01:09:05,439
level of the entire workforce.

1527
01:09:05,439 --> 01:09:08,000
STEPH: And increasing
social labor protections.

1528
01:09:08,000 --> 01:09:09,875
WILLIAM BONVILLIAN: Yes,
and we did that too.

1529
01:09:11,890 --> 01:09:14,115
MARTIN: Yeah, that is a
fundamental assumption

1530
01:09:14,115 --> 01:09:16,240
that I think is pretty
interesting-- the assumption

1531
01:09:16,240 --> 01:09:17,698
that there should
be a middle class

1532
01:09:17,698 --> 01:09:19,840
or that it's natural
for there to be.

1533
01:09:19,840 --> 01:09:22,680
Or if that's always
going to be a case.

1534
01:09:22,680 --> 01:09:24,779
Not to say it shouldn't
be or it should be.

1535
01:09:24,779 --> 01:09:26,196
Just it's interesting
that we have

1536
01:09:26,196 --> 01:09:28,960
the assumption that the
fundamentals have always shown

1537
01:09:28,960 --> 01:09:30,640
that that should be a thing.

1538
01:09:30,640 --> 01:09:32,182
WILLIAM BONVILLIAN:
I think I'd argue

1539
01:09:32,182 --> 01:09:35,547
that that's pretty fundamental
to the ethos of the country.

1540
01:09:35,547 --> 01:09:36,880
STEPH: The American model, yeah.

1541
01:09:36,880 --> 01:09:40,960
WILLIAM BONVILLIAN: Overall,
immigration occurred here,

1542
01:09:40,960 --> 01:09:43,540
and social classes
were significantly

1543
01:09:43,540 --> 01:09:47,140
reduced in the United States
from the period of founding

1544
01:09:47,140 --> 01:09:49,479
until the latter part
of the 19th century

1545
01:09:49,479 --> 01:09:51,776
when industrialization took off.

1546
01:09:51,776 --> 01:09:53,859
But that it was a pretty
strong middle-class base.

1547
01:09:53,859 --> 01:09:56,860
And that talent base, that
involved base, that citizen

1548
01:09:56,860 --> 01:09:59,530
base, is pretty
key to the effect--

1549
01:09:59,530 --> 01:10:02,800
the ability of
democracy to function.

1550
01:10:02,800 --> 01:10:05,020
And what happens
in democracy when

1551
01:10:05,020 --> 01:10:07,240
you start to really
split the society

1552
01:10:07,240 --> 01:10:09,962
and significant
inequality occurs?

1553
01:10:09,962 --> 01:10:11,920
I'm worried about some
of this signaling system

1554
01:10:11,920 --> 01:10:13,140
we're getting on
that at the moment.

1555
01:10:13,140 --> 01:10:13,280
MARTIN: Yeah.

1556
01:10:13,280 --> 01:10:15,197
I think it'd be interesting
to do a comparison

1557
01:10:15,197 --> 01:10:16,705
to another time in history--

1558
01:10:16,705 --> 01:10:18,747
STEPH: Like feudalism?

1559
01:10:18,747 --> 01:10:20,080
MARTIN: Well, yeah, just to go--

1560
01:10:20,080 --> 01:10:24,163
I mean, [INAUDIBLE] from there,
because we definitely are

1561
01:10:24,163 --> 01:10:25,330
living in interesting times.

1562
01:10:28,760 --> 01:10:31,600
AUDIENCE: But intuitively,
after learning economics

1563
01:10:31,600 --> 01:10:35,260
for several years, I think
raising the labor mobility is

1564
01:10:35,260 --> 01:10:38,950
probably one of the solutions.

1565
01:10:38,950 --> 01:10:44,590
So how do you guys feel about
how effective would this be

1566
01:10:44,590 --> 01:10:46,730
and how feasible would it be?

1567
01:10:46,730 --> 01:10:48,310
MARTIN: Labor mobility?

1568
01:10:48,310 --> 01:10:50,170
AUDIENCE: Raising
labor mobilities.

1569
01:10:50,170 --> 01:10:51,210
MAX: Doesn't that increase
inefficiency, though?

1570
01:10:51,210 --> 01:10:52,990
WILLIAM BONVILLIAN: Well,
we've decreased labor mobility

1571
01:10:52,990 --> 01:10:54,407
in the United
States significantly

1572
01:10:54,407 --> 01:10:55,580
in the last several decades.

1573
01:10:55,580 --> 01:10:57,432
RASHEED: Wait, define
labor mobility--

1574
01:10:57,432 --> 01:10:59,390
WILLIAM BONVILLIAN: The
ability to move up in--

1575
01:10:59,390 --> 01:11:01,390
AUDIENCE: Yeah, once they
lost their employment,

1576
01:11:01,390 --> 01:11:03,430
they can move to
find new employments

1577
01:11:03,430 --> 01:11:04,710
with a new set of skills.

1578
01:11:04,710 --> 01:11:05,380
RASHEED: Got it.

1579
01:11:05,380 --> 01:11:06,463
WILLIAM BONVILLIAN: Right.

1580
01:11:06,463 --> 01:11:08,950
But what happens in these
failing industrial economies

1581
01:11:08,950 --> 01:11:10,690
is that people--

1582
01:11:10,690 --> 01:11:13,840
homeownership is typically
a family's key asset.

1583
01:11:13,840 --> 01:11:17,290
If the community is failing
because of industrial decline,

1584
01:11:17,290 --> 01:11:18,520
that asset collapses.

1585
01:11:18,520 --> 01:11:21,910
It's much harder for people
to get out at that point.

1586
01:11:21,910 --> 01:11:24,450
So there's been
less labor mobility,

1587
01:11:24,450 --> 01:11:28,750
and interestingly, less job
churn in the United States

1588
01:11:28,750 --> 01:11:29,740
in the last decade.

1589
01:11:29,740 --> 01:11:32,230
And then there's
social mobility,

1590
01:11:32,230 --> 01:11:35,610
which arguably has been
pretty key to this country.

1591
01:11:35,610 --> 01:11:37,785
In other words, your
ability to improve yourself,

1592
01:11:37,785 --> 01:11:39,660
which has always been
an operating assumption

1593
01:11:39,660 --> 01:11:40,570
in the United States.

1594
01:11:40,570 --> 01:11:43,630
You could always do better
than your parents, right?

1595
01:11:43,630 --> 01:11:46,180
And that may be coming
to an end, as well.

1596
01:11:46,180 --> 01:11:48,040
That may be significantly
forward closed.

1597
01:11:48,040 --> 01:11:49,915
MARTIN: They've shown
that this generation is

1598
01:11:49,915 --> 01:11:52,110
worse off in real numbers
than their parents were.

1599
01:11:52,110 --> 01:11:54,592
Or they're less likely
to do what they do.

1600
01:11:54,592 --> 01:11:56,800
WILLIAM BONVILLIAN: Fortunately,
you have more stuff.

1601
01:11:56,800 --> 01:11:57,970
That's the one really--

1602
01:11:57,970 --> 01:11:59,520
MARTIN: Yeah, there's
a saying where,

1603
01:11:59,520 --> 01:12:01,062
if you could choose
to be alive today

1604
01:12:01,062 --> 01:12:02,910
and be poor, or
be a billionaire--

1605
01:12:02,910 --> 01:12:04,270
WILLIAM BONVILLIAN: We may
be poor, but we got iPhones.

1606
01:12:04,270 --> 01:12:05,728
MARTIN: --like a
multi-millionaire.

1607
01:12:05,728 --> 01:12:07,700
Yeah, it's like you're
a rich-poor, right?

1608
01:12:07,700 --> 01:12:08,650
It's like, yeah, I
can't afford anything.

1609
01:12:08,650 --> 01:12:10,020
But I got my phone, I got--

1610
01:12:10,020 --> 01:12:11,768
MAX: I got beams.

1611
01:12:11,768 --> 01:12:13,810
STEPH: Actually, interestingly,
in terms of memes

1612
01:12:13,810 --> 01:12:16,420
representing your culture,
I saw a meme last night

1613
01:12:16,420 --> 01:12:18,808
that said, why do you think
I post so many selfies?

1614
01:12:18,808 --> 01:12:21,350
I don't have a car, I don't have
a house to post pictures of,

1615
01:12:21,350 --> 01:12:23,450
so I'm just going to
post photos of my face.

1616
01:12:23,450 --> 01:12:25,318
And I think that's really--

1617
01:12:25,318 --> 01:12:26,110
MAX: That's so sad.

1618
01:12:26,110 --> 01:12:27,020
STEPH: No, I think
that's representative

1619
01:12:27,020 --> 01:12:28,930
of our social
opportunities, right?

1620
01:12:28,930 --> 01:12:29,515
AUDIENCE: And you
know that millionaire

1621
01:12:29,515 --> 01:12:32,148
who was complaining that the
reason that millennials don't

1622
01:12:32,148 --> 01:12:33,940
have houses is because
they're spending all

1623
01:12:33,940 --> 01:12:35,250
their money on avocado toast.

1624
01:12:35,250 --> 01:12:37,610
[LAUGHTER]

1625
01:12:38,300 --> 01:12:39,550
WILLIAM BONVILLIAN: All right.

1626
01:12:39,550 --> 01:12:41,050
So I'm going to
take the opportunity

1627
01:12:41,050 --> 01:12:43,930
to do a quick wrap up
of just this class,

1628
01:12:43,930 --> 01:12:46,030
because we did the wrap
up of the overall class

1629
01:12:46,030 --> 01:12:47,320
at the outset.

1630
01:12:47,320 --> 01:12:49,030
But just very briefly here--

1631
01:12:53,330 --> 01:13:00,200
Brynjolfsson and McAfee tell
us that the advent of IT

1632
01:13:00,200 --> 01:13:05,126
is going to be
significantly disruptive

1633
01:13:05,126 --> 01:13:09,890
in the foreseeable future, in
the short to mid-term future.

1634
01:13:09,890 --> 01:13:12,140
And it's going to create
significant technological job

1635
01:13:12,140 --> 01:13:12,950
displacement.

1636
01:13:12,950 --> 01:13:16,610
And David Autor
in his piece "Why

1637
01:13:16,610 --> 01:13:18,980
Do We Still Have So Many Jobs?"

1638
01:13:18,980 --> 01:13:21,980
responds with a
long-standing set

1639
01:13:21,980 --> 01:13:24,140
of thoughtful
economic arguments,

1640
01:13:24,140 --> 01:13:27,830
arguing that we can't just
look at the displaced jobs.

1641
01:13:27,830 --> 01:13:30,830
We also have to look at the
new complementary jobs that

1642
01:13:30,830 --> 01:13:34,790
are coming about and the net
increase in new technologies

1643
01:13:34,790 --> 01:13:36,950
and production
that will accompany

1644
01:13:36,950 --> 01:13:39,290
these technological
changes, too.

1645
01:13:39,290 --> 01:13:44,660
They create a ground
condition that

1646
01:13:44,660 --> 01:13:48,680
may make this shift
more manageable.

1647
01:13:48,680 --> 01:13:52,790
And Autor also talks
about how our economy--

1648
01:13:52,790 --> 01:13:55,640
he looks at
inequality and ties it

1649
01:13:55,640 --> 01:14:00,170
to education and
education attainment.

1650
01:14:00,170 --> 01:14:05,900
And he argues overall that
this inequality problem

1651
01:14:05,900 --> 01:14:09,567
is tied to our education
levels, education

1652
01:14:09,567 --> 01:14:10,650
attainment in our society.

1653
01:14:10,650 --> 01:14:13,550
It's going to require
significant upskilling

1654
01:14:13,550 --> 01:14:15,440
of our population.

1655
01:14:15,440 --> 01:14:17,630
And then David
Mindell takes us back

1656
01:14:17,630 --> 01:14:20,350
to the, what's the future of
working going to look like?

1657
01:14:20,350 --> 01:14:23,570
And takes a deep dive into what
most people view as the most

1658
01:14:23,570 --> 01:14:25,820
threatening
territory-- robotics--

1659
01:14:25,820 --> 01:14:29,720
and argues that, yes, there
will be technological job

1660
01:14:29,720 --> 01:14:32,390
displacement, but the
relationship with robotics

1661
01:14:32,390 --> 01:14:34,610
is a complementary
one, that there's

1662
01:14:34,610 --> 01:14:36,590
going to be a human
machine symbiosis.

1663
01:14:36,590 --> 01:14:39,380
It's going to be a more
complicated picture than just

1664
01:14:39,380 --> 01:14:41,450
straight technological
job displacement.

1665
01:14:41,450 --> 01:14:45,410
There will be new opportunities
for people in this mix.

1666
01:14:45,410 --> 01:14:49,280
And then reading from that
chapter of the upcoming book

1667
01:14:49,280 --> 01:14:55,063
that I'm doing
with Peter Singer,

1668
01:14:55,063 --> 01:14:56,480
looking through
the literature, we

1669
01:14:56,480 --> 01:15:00,830
conclude that the jobless
future is not on us today,

1670
01:15:00,830 --> 01:15:04,040
and the low productivity
rates and low investment rates

1671
01:15:04,040 --> 01:15:08,390
suggest that it's
not evolving rapidly.

1672
01:15:08,390 --> 01:15:10,700
So we may have a
little time here.

1673
01:15:10,700 --> 01:15:13,400
And then, meanwhile, we
do have this deep problem

1674
01:15:13,400 --> 01:15:15,350
with a low growth rate,
a low productivity

1675
01:15:15,350 --> 01:15:18,950
rate, a low capital investment
rate, which is creating secular

1676
01:15:18,950 --> 01:15:20,695
stagnation, in Summers' terms.

1677
01:15:20,695 --> 01:15:22,070
And that's something
that we need

1678
01:15:22,070 --> 01:15:26,090
to get on, because that's
the immediate future.

1679
01:15:26,090 --> 01:15:29,930
And then, overall, this general
upskilling of the workforce

1680
01:15:29,930 --> 01:15:31,520
serves both purposes.

1681
01:15:31,520 --> 01:15:33,910
So let's get on with that.