1
00:00:00,240 --> 00:00:05,159
Speaker 1: From self driving cars to robot powered factories. Artificial intelligence

2
00:00:05,240 --> 00:00:09,360
Speaker 1: is taking over significant pieces of the global economy. This

3
00:00:09,440 --> 00:00:12,880
Speaker 1: is great for the businesses embracing AI, but there is

4
00:00:12,880 --> 00:00:16,599
Speaker 1: a downside. More robots in the workforce also means more

5
00:00:16,720 --> 00:00:20,560
Speaker 1: people losing their jobs to computers. So how bad will

6
00:00:20,600 --> 00:00:23,919
Speaker 1: the robot revolution be and how will it reshape the

7
00:00:23,960 --> 00:00:37,200
Speaker 1: global economy? Welcome to Benchmark. I'm Scott Landman and economics

8
00:00:37,280 --> 00:00:40,840
Speaker 1: editor with Bloomberg News in Washington. Returning as a guest

9
00:00:40,880 --> 00:00:44,199
Speaker 1: co host is my colleague Chris Content. He's a reporter

10
00:00:44,320 --> 00:00:47,239
Speaker 1: covering the Federal Reserve and u S economy also here

11
00:00:47,280 --> 00:00:49,960
Speaker 1: in d C. Chris, glad to have you back, Happy

12
00:00:50,000 --> 00:00:52,479
Speaker 1: to be here. So this week we're talking about the

13
00:00:52,600 --> 00:00:55,840
Speaker 1: rise of the robots and how they will impact the economy.

14
00:00:56,280 --> 00:00:59,360
Speaker 1: As someone who writes about the Federal Reserve, Chris, what

15
00:00:59,600 --> 00:01:02,680
Speaker 1: interest you and AI? Well, I can tell you, Scott,

16
00:01:02,760 --> 00:01:06,640
Speaker 1: that there is a great capacity actually of artificial intelligence

17
00:01:06,680 --> 00:01:11,560
Speaker 1: to transform one of the biggest jobs of central banks.

18
00:01:11,600 --> 00:01:15,360
Speaker 1: As you know, central banks are trying to regulate economies,

19
00:01:15,480 --> 00:01:18,160
Speaker 1: keep them in that Goldilock zone, and they do this

20
00:01:18,360 --> 00:01:22,440
Speaker 1: through short term interest rates. But that tool has a

21
00:01:22,560 --> 00:01:25,080
Speaker 1: lag to it. It takes a good six st eighteen

22
00:01:25,200 --> 00:01:28,880
Speaker 1: months before a move and interest rates has an impact

23
00:01:29,280 --> 00:01:32,920
Speaker 1: in the real economy. So these central bankers they've got

24
00:01:32,959 --> 00:01:36,319
Speaker 1: to forecast what what's the economy gonna be like a

25
00:01:36,440 --> 00:01:39,760
Speaker 1: year from now, eighteen months from now. But they're not

26
00:01:40,080 --> 00:01:43,120
Speaker 1: very good at forecasting the future when it comes to

27
00:01:43,200 --> 00:01:49,120
Speaker 1: things like inflation, unemployment, GDP growth. So some central banks

28
00:01:49,160 --> 00:01:53,040
Speaker 1: are starting to turn for a little help to artificial

29
00:01:53,080 --> 00:01:58,040
Speaker 1: intelligence because artificial intelligence can be used quite effectively at

30
00:01:58,400 --> 00:02:03,280
Speaker 1: spotting patterns in past events and then using that to

31
00:02:03,440 --> 00:02:05,480
Speaker 1: predict the future. So it's an area it's going to

32
00:02:05,560 --> 00:02:07,920
Speaker 1: take a while, but it's an area of great promise

33
00:02:08,280 --> 00:02:13,000
Speaker 1: for economics. It's not just economics, but it's business and

34
00:02:13,560 --> 00:02:17,720
Speaker 1: the broader economy as a whole. Forecasting prediction, it's it's

35
00:02:17,720 --> 00:02:20,880
Speaker 1: a really big part of what AI is meant to do.

36
00:02:21,360 --> 00:02:24,080
Speaker 1: And our guest is here to talk about that today.

37
00:02:24,440 --> 00:02:27,400
Speaker 1: His name is Joshua Gains and he's a professor at

38
00:02:27,440 --> 00:02:31,680
Speaker 1: the University of Toronto's Rotman School of Management. He's one

39
00:02:31,680 --> 00:02:35,120
Speaker 1: of the authors of a brand new book called Prediction Machines,

40
00:02:35,320 --> 00:02:39,760
Speaker 1: The Simple Economics of Artificial Intelligence, just published by Harvard

41
00:02:39,800 --> 00:02:43,440
Speaker 1: Business Review Press. Joshua, thanks for joining us. Good to

42
00:02:43,440 --> 00:02:46,359
Speaker 1: be here. Thanks. So we've already sort of alluded to this,

43
00:02:46,400 --> 00:02:50,680
Speaker 1: but why is the book called Prediction Machines. Well, it's

44
00:02:50,880 --> 00:02:55,400
Speaker 1: uh called that and not some more interesting title such as, um,

45
00:02:55,440 --> 00:02:59,120
Speaker 1: you know you're wonderful, great new robot and how your

46
00:02:59,120 --> 00:03:02,280
Speaker 1: life is going to get better, simply because the recent

47
00:03:02,320 --> 00:03:07,560
Speaker 1: developments in artificial intelligence are not about completely replacing human

48
00:03:07,600 --> 00:03:11,560
Speaker 1: intelligence per se, but actually all of being about one thing,

49
00:03:11,720 --> 00:03:14,960
Speaker 1: and that is prediction. That is taking a whole lot

50
00:03:14,960 --> 00:03:18,160
Speaker 1: of information that you do have and converting it into

51
00:03:18,240 --> 00:03:21,919
Speaker 1: information you do not have. That's very, very different from

52
00:03:21,960 --> 00:03:24,840
Speaker 1: something that does all of your choices for you and

53
00:03:24,880 --> 00:03:28,840
Speaker 1: things like that. Uh, it really really just does prediction.

54
00:03:29,520 --> 00:03:33,400
Speaker 1: And when we talk about this prediction function and getting

55
00:03:33,440 --> 00:03:36,320
Speaker 1: all this information, can you bring us into the real

56
00:03:36,360 --> 00:03:39,920
Speaker 1: world here and what kinds of industries or jobs would

57
00:03:39,960 --> 00:03:42,680
Speaker 1: be most likely to or have the most room to

58
00:03:42,760 --> 00:03:47,880
Speaker 1: benefit from this kind of additional knowledge. Well, prediction is

59
00:03:47,920 --> 00:03:51,320
Speaker 1: something and from your business that you're in, you think

60
00:03:51,360 --> 00:03:55,440
Speaker 1: of it mainly about forecasting, for instance, economic variables and

61
00:03:55,480 --> 00:03:58,560
Speaker 1: things like that, and something of course I've worried about

62
00:03:58,600 --> 00:04:02,720
Speaker 1: as well. But what's really interesting about these new developments

63
00:04:02,960 --> 00:04:07,000
Speaker 1: is that they highlight problems that turn out to be

64
00:04:07,080 --> 00:04:11,600
Speaker 1: prediction problems. For instance, the whole issue of having a

65
00:04:11,640 --> 00:04:14,520
Speaker 1: digital image served up to you and knowing what's in

66
00:04:14,560 --> 00:04:17,960
Speaker 1: it is a prediction problem. Invariably, when you get an

67
00:04:17,960 --> 00:04:20,839
Speaker 1: image from the Internet, the label that's being attached to

68
00:04:20,880 --> 00:04:23,880
Speaker 1: it is the label that human would attach to it.

69
00:04:24,279 --> 00:04:26,960
Speaker 1: And so basically what Google are doing when you're searching

70
00:04:26,960 --> 00:04:31,080
Speaker 1: for a picture is predicting which pictures correspond to that label.

71
00:04:31,600 --> 00:04:34,680
Speaker 1: So that's a form of prediction. And it turns out

72
00:04:34,720 --> 00:04:40,760
Speaker 1: that that and language translation and understanding machines, understanding human speech,

73
00:04:40,800 --> 00:04:46,880
Speaker 1: and even self driving cars are all mainly a prediction problem,

74
00:04:46,960 --> 00:04:50,120
Speaker 1: and so that is where this new artificial intelligence is

75
00:04:50,160 --> 00:04:54,800
Speaker 1: being implemented. You wouldn't normally call that prediction problems. We

76
00:04:54,839 --> 00:04:58,200
Speaker 1: normally think about forecasting the weather or forecasting an earthquake

77
00:04:58,279 --> 00:05:02,200
Speaker 1: or something, but basically prediction is all around us. Joshua,

78
00:05:02,320 --> 00:05:04,719
Speaker 1: I think it's clear that this can have all sorts

79
00:05:04,720 --> 00:05:08,000
Speaker 1: of benefits in our lives and in the economy, but

80
00:05:08,040 --> 00:05:11,440
Speaker 1: I can also have some downsides, and particularly how how

81
00:05:11,520 --> 00:05:16,000
Speaker 1: these things are distributed throughout our society. Maybe a big

82
00:05:16,080 --> 00:05:19,160
Speaker 1: question mark first about the benefits. Where where do you

83
00:05:19,200 --> 00:05:23,760
Speaker 1: see that societally and also economically. So I think the

84
00:05:23,800 --> 00:05:28,240
Speaker 1: benefits come from wherever you think about where would it

85
00:05:28,240 --> 00:05:32,279
Speaker 1: be good to know things with greater certainty? And so

86
00:05:32,320 --> 00:05:36,120
Speaker 1: any kind of decisions that you're doing under uncertainty are

87
00:05:36,120 --> 00:05:40,760
Speaker 1: going to benefit from having better prediction so that you

88
00:05:40,800 --> 00:05:45,120
Speaker 1: can imagine being applied. For instance, you're trying to manage inventory.

89
00:05:45,240 --> 00:05:48,200
Speaker 1: If you've got better predictions regarding demand you're going to face,

90
00:05:48,240 --> 00:05:50,400
Speaker 1: you're going to be able to manage that inventory better.

91
00:05:50,760 --> 00:05:53,839
Speaker 1: You're going to be able to correctly adjust your behavior

92
00:05:53,880 --> 00:05:59,360
Speaker 1: to prevent shortfalls or worse, to prevent surpluses that end

93
00:05:59,440 --> 00:06:02,839
Speaker 1: up to over stocking your inventory. Those are the sorts

94
00:06:02,880 --> 00:06:07,119
Speaker 1: of places where prediction machines are going to work quite well. Uh.

95
00:06:07,120 --> 00:06:11,640
Speaker 1: And so basically anywhere where there's decision making being done

96
00:06:12,000 --> 00:06:15,839
Speaker 1: and there's uncertainty, there is room for better prediction, and

97
00:06:15,880 --> 00:06:18,760
Speaker 1: the machines might serve that up right, And that sounds,

98
00:06:18,800 --> 00:06:21,640
Speaker 1: of course like it will make our companies more efficient,

99
00:06:21,680 --> 00:06:26,280
Speaker 1: more productive, but also I think will disrupt how companies

100
00:06:26,320 --> 00:06:30,440
Speaker 1: operate and may disrupt people's lives. Do you think this

101
00:06:30,520 --> 00:06:33,119
Speaker 1: time is going to be any different. We've had lots

102
00:06:33,160 --> 00:06:37,719
Speaker 1: of technology disruption in our economy over the decades, in fact,

103
00:06:37,720 --> 00:06:41,679
Speaker 1: in centuries, um is the pace of change these days

104
00:06:41,720 --> 00:06:45,839
Speaker 1: going to be more disruptive and more problematic to adjust to.

105
00:06:46,520 --> 00:06:49,200
Speaker 1: I think there's a chance, as usual, whenever you've got

106
00:06:49,200 --> 00:06:54,640
Speaker 1: a very large and radical innovation occurring and being adopted,

107
00:06:55,000 --> 00:06:58,679
Speaker 1: there is the potential for disruptive change. But the hard

108
00:06:58,720 --> 00:07:01,400
Speaker 1: part is trying to predict exactly where that will be.

109
00:07:02,520 --> 00:07:05,000
Speaker 1: When we look back and we think about the rise

110
00:07:05,000 --> 00:07:08,400
Speaker 1: of the automobile, it's no surprise that we also point

111
00:07:08,440 --> 00:07:13,160
Speaker 1: to horses as being the disruptive workers in that disrupted

112
00:07:13,200 --> 00:07:17,040
Speaker 1: workers in that equation. When it comes to things like

113
00:07:17,520 --> 00:07:21,480
Speaker 1: better prediction, however, it's far subtler um. Some of the

114
00:07:21,480 --> 00:07:25,320
Speaker 1: discussion that's going around is saying, well, this is the

115
00:07:25,360 --> 00:07:27,840
Speaker 1: first Why is this time different? It's because it's the

116
00:07:27,880 --> 00:07:32,160
Speaker 1: first time you really have machines taking over cognitive tasks.

117
00:07:32,680 --> 00:07:35,720
Speaker 1: Well that's not true. We've had machines take over cognitive

118
00:07:35,760 --> 00:07:39,920
Speaker 1: task with computers, and you know, we seem to most

119
00:07:39,920 --> 00:07:42,600
Speaker 1: of us seem to be still be gainfully employed, and

120
00:07:42,640 --> 00:07:47,120
Speaker 1: the adjustments that took place have been worked out uh,

121
00:07:47,440 --> 00:07:49,920
Speaker 1: this time, you know, Okay, the computers are doing much

122
00:07:50,000 --> 00:07:53,480
Speaker 1: much more. They're doing much more in terms of thought. Yes,

123
00:07:53,600 --> 00:07:56,240
Speaker 1: But as we identify in the book, the one thing

124
00:07:56,240 --> 00:08:00,280
Speaker 1: that computers can't do is set goals. You still have

125
00:08:00,440 --> 00:08:04,400
Speaker 1: to have a human what we call human judgment, to

126
00:08:04,560 --> 00:08:07,560
Speaker 1: set the goals the trade offs. No prediction is going

127
00:08:07,600 --> 00:08:09,600
Speaker 1: to be perfect, so you have to work out what

128
00:08:10,320 --> 00:08:12,640
Speaker 1: how you're going to stomach errors and things like that.

129
00:08:13,360 --> 00:08:17,440
Speaker 1: Those are still roles for people to get into. It's

130
00:08:17,480 --> 00:08:21,720
Speaker 1: only where better prediction was like the final thing towards

131
00:08:21,800 --> 00:08:25,880
Speaker 1: getting full automation that you might see jobs actually replaced.

132
00:08:26,840 --> 00:08:31,400
Speaker 1: Speaking of not getting to full automation, one interesting example

133
00:08:31,520 --> 00:08:34,240
Speaker 1: you discussed a little bit in the book is that

134
00:08:34,400 --> 00:08:39,240
Speaker 1: of doctors and how their jobs might change under advances

135
00:08:39,280 --> 00:08:43,400
Speaker 1: in AI, in terms of they just won't have to

136
00:08:43,520 --> 00:08:46,600
Speaker 1: do the diagnoses as much anymore. The computer is going

137
00:08:46,640 --> 00:08:48,200
Speaker 1: to do it for them, and their jobs are going

138
00:08:48,240 --> 00:08:51,600
Speaker 1: to substantially change. Can you talk about that a bit well,

139
00:08:52,040 --> 00:08:55,640
Speaker 1: basically in a conference a couple of years ago. Jeff Hinton,

140
00:08:55,960 --> 00:09:00,000
Speaker 1: who is one of the pioneers of the new development

141
00:09:00,000 --> 00:09:02,600
Speaker 1: it's in AI. He now it works part time at Google.

142
00:09:03,240 --> 00:09:06,120
Speaker 1: He basically said to the conference, well, I think we

143
00:09:06,120 --> 00:09:10,199
Speaker 1: should stop training radiologists. Now, what he meant by that

144
00:09:10,440 --> 00:09:14,200
Speaker 1: was his view of what a radiologist does is look

145
00:09:14,280 --> 00:09:18,160
Speaker 1: at images and then decide, you know, is there a

146
00:09:18,240 --> 00:09:23,280
Speaker 1: problem or not that requires further treatment. Well, obviously, prediction

147
00:09:23,320 --> 00:09:26,800
Speaker 1: machines have the capability and have been demonstrated in some

148
00:09:26,840 --> 00:09:29,840
Speaker 1: settings to be far superior to people looking at those

149
00:09:29,880 --> 00:09:33,960
Speaker 1: pictures and identifying exactly what's going on with a far

150
00:09:34,040 --> 00:09:36,960
Speaker 1: greater degree of accuracy. So if that's your view of

151
00:09:36,960 --> 00:09:40,680
Speaker 1: what a radiologist does, well it sounds like curtains for them.

152
00:09:41,280 --> 00:09:44,240
Speaker 1: But actually radiologists have been dealing with these sorts of

153
00:09:44,240 --> 00:09:47,640
Speaker 1: issues for fifty years and they're quite aware of them.

154
00:09:47,640 --> 00:09:52,000
Speaker 1: That has technology improves their jobs change. What happens in

155
00:09:52,080 --> 00:09:55,320
Speaker 1: terms of radiologists is it's not some a simple choice.

156
00:09:56,280 --> 00:09:58,800
Speaker 1: You get a prediction of something and you know exactly

157
00:09:58,840 --> 00:10:01,199
Speaker 1: what to do. You're right. If that was the case,

158
00:10:01,240 --> 00:10:03,120
Speaker 1: then anyone could just look it up in a manual

159
00:10:03,480 --> 00:10:06,920
Speaker 1: and you wouldn't need a radiologist. But invariably there are

160
00:10:06,960 --> 00:10:11,520
Speaker 1: other factors, other criteria, and in particular the personal dimension

161
00:10:11,720 --> 00:10:15,760
Speaker 1: of the patient situations that will also impact on the

162
00:10:15,800 --> 00:10:19,560
Speaker 1: what the treatment decision actually is. So prediction is an

163
00:10:19,600 --> 00:10:23,240
Speaker 1: important input into that, but there are other factors going on,

164
00:10:23,800 --> 00:10:26,679
Speaker 1: and we're a long way off being able to automate

165
00:10:26,760 --> 00:10:31,120
Speaker 1: all of that. And thus far, the evidence is that

166
00:10:31,400 --> 00:10:34,319
Speaker 1: you make the radiologists actually better at their jobs by

167
00:10:34,360 --> 00:10:37,000
Speaker 1: having these prediction machines. They are able to act with

168
00:10:37,080 --> 00:10:41,120
Speaker 1: more certainty and therefore able to come up with more

169
00:10:41,120 --> 00:10:45,200
Speaker 1: confident recommendations. This allows them to save some time and

170
00:10:45,240 --> 00:10:48,400
Speaker 1: save some other things, and develop other skills, and may

171
00:10:48,520 --> 00:10:53,079
Speaker 1: change the allocation of tasks between radiologists and other medical

172
00:10:53,080 --> 00:10:57,320
Speaker 1: practitioners quite a bit. So I don't necessarily see it

173
00:10:57,480 --> 00:11:00,800
Speaker 1: is obvious that those jobs are going to be disrupted

174
00:11:00,800 --> 00:11:04,400
Speaker 1: as quickly as some of the engineers do. What kind

175
00:11:04,400 --> 00:11:08,760
Speaker 1: of jobs, Josh, would you say are the least vulnerable

176
00:11:08,960 --> 00:11:12,880
Speaker 1: to being replaced or even disrupted by this kind of technology.

177
00:11:13,480 --> 00:11:16,679
Speaker 1: Well that's a really interesting question. You know. We tend to,

178
00:11:17,120 --> 00:11:20,320
Speaker 1: when confronted with this thing, point to all the important

179
00:11:20,360 --> 00:11:24,319
Speaker 1: things that we do and how it can't be replaced

180
00:11:24,360 --> 00:11:27,959
Speaker 1: by a machine. For instance, some of the jobs that

181
00:11:28,080 --> 00:11:31,280
Speaker 1: people talk about as being hard to be replaced by

182
00:11:31,360 --> 00:11:36,640
Speaker 1: machines are ones that require emotional input. Interestingly, Danny Khneman

183
00:11:36,840 --> 00:11:41,600
Speaker 1: at a conference here just last year. He's the Nobel

184
00:11:41,679 --> 00:11:44,920
Speaker 1: Prize winning economist who has been responsible for behavior or

185
00:11:44,960 --> 00:11:49,200
Speaker 1: economics and thinking about decision making and judgment in the

186
00:11:49,240 --> 00:11:54,120
Speaker 1: context of psychology, and his answer was equivocal. He sees

187
00:11:54,200 --> 00:11:57,720
Speaker 1: humans as ultimately very flawed and doesn't see any reason

188
00:11:57,800 --> 00:12:01,360
Speaker 1: why machines wouldn't take over. His view was, do you

189
00:12:01,440 --> 00:12:05,560
Speaker 1: really want your care to be managed by disinterested children

190
00:12:06,040 --> 00:12:09,080
Speaker 1: when you've become elderly, or would you rather have a

191
00:12:09,160 --> 00:12:12,319
Speaker 1: robot who's been trained to work out exactly what your

192
00:12:12,360 --> 00:12:15,160
Speaker 1: needs are. So it's very hard to work out what

193
00:12:15,360 --> 00:12:18,120
Speaker 1: is safe and what isn't. All we can say right

194
00:12:18,160 --> 00:12:21,640
Speaker 1: now is that the tools in their current instantiation and

195
00:12:21,720 --> 00:12:23,840
Speaker 1: what we're going to see probably over the next five

196
00:12:23,920 --> 00:12:27,240
Speaker 1: years and maybe ten years, is all about that prediction function,

197
00:12:27,520 --> 00:12:30,680
Speaker 1: which leaves a lot of room open for people. Speaking

198
00:12:30,679 --> 00:12:33,880
Speaker 1: of jobs that are likely safe from AI disruption, I

199
00:12:33,880 --> 00:12:40,079
Speaker 1: think Chris will appreciate being in Washington, d C. Federal government, Congress, etcetera.

200
00:12:40,160 --> 00:12:43,120
Speaker 1: Those will probably be pretty safe from disruption for a

201
00:12:43,160 --> 00:12:46,160
Speaker 1: long time. So if we ever get some disruption that sector,

202
00:12:46,280 --> 00:12:49,560
Speaker 1: that would be something definitely to watch. But josh I

203
00:12:49,600 --> 00:12:53,040
Speaker 1: wanted to turn to another issue that you discuss when

204
00:12:53,120 --> 00:12:56,520
Speaker 1: when you're talking about economic impact in the book. Uh,

205
00:12:56,559 --> 00:13:00,760
Speaker 1: it's really intriguing that you talk about inequality and how

206
00:13:00,880 --> 00:13:05,960
Speaker 1: that would evolve if AI were to take hold in

207
00:13:06,040 --> 00:13:10,280
Speaker 1: the economy. Why could that get worse with the rise

208
00:13:10,400 --> 00:13:14,160
Speaker 1: of AI. Well, it's you know, as usual, it's difficult

209
00:13:14,160 --> 00:13:17,800
Speaker 1: to forecast these things. You never should trust economis on

210
00:13:18,160 --> 00:13:21,280
Speaker 1: on those sorts of big trend forecasts. But the concern

211
00:13:21,360 --> 00:13:26,720
Speaker 1: that we have is that certain skills become more valuable

212
00:13:26,840 --> 00:13:30,560
Speaker 1: valuable and other skills become less so. And one of

213
00:13:30,600 --> 00:13:33,640
Speaker 1: the problems is if you wanted to be the sort

214
00:13:33,679 --> 00:13:37,559
Speaker 1: of person who could take advantage of AI, it's invariably

215
00:13:37,880 --> 00:13:42,079
Speaker 1: not going to be a skill that is like routine. Necessarily,

216
00:13:42,720 --> 00:13:44,600
Speaker 1: once you know, once you have the sort of job

217
00:13:44,600 --> 00:13:46,640
Speaker 1: where you can look at a prediction and you know

218
00:13:46,679 --> 00:13:50,280
Speaker 1: exactly what to do, your place in that is devalued.

219
00:13:50,880 --> 00:13:53,640
Speaker 1: On the other hand, there are some situations where it's

220
00:13:53,679 --> 00:13:55,440
Speaker 1: going to take a bit of art to understand what

221
00:13:55,520 --> 00:13:59,000
Speaker 1: the prediction really is. We don't tend to think of

222
00:13:59,360 --> 00:14:02,800
Speaker 1: machine learn in computer sciences art, but there are so

223
00:14:02,840 --> 00:14:06,480
Speaker 1: many variables that the engineers have to control and think

224
00:14:06,480 --> 00:14:09,719
Speaker 1: about that. It does tend to have that quality to it,

225
00:14:10,120 --> 00:14:13,840
Speaker 1: and similarly, how we use those tools also has a

226
00:14:13,880 --> 00:14:16,720
Speaker 1: bit of an artistic flare to it. And by that

227
00:14:16,840 --> 00:14:20,440
Speaker 1: I mean that it's not a dents sure why someone

228
00:14:20,480 --> 00:14:24,520
Speaker 1: who's more productive is more productive, but somehow they get

229
00:14:24,560 --> 00:14:27,240
Speaker 1: it more and so the people who are able to

230
00:14:27,280 --> 00:14:30,120
Speaker 1: do that are probably going to fare better. The one

231
00:14:30,160 --> 00:14:33,280
Speaker 1: concern we have is when I'm a person who's using

232
00:14:33,280 --> 00:14:35,920
Speaker 1: an AI tool, I can use it at a much

233
00:14:36,080 --> 00:14:41,480
Speaker 1: larger scale and make decisions for many more across a

234
00:14:41,560 --> 00:14:45,040
Speaker 1: greater domain. And in that regard, I'm sort of made

235
00:14:45,040 --> 00:14:48,760
Speaker 1: a superhuman. But you can only have so many superhumans,

236
00:14:48,800 --> 00:14:52,520
Speaker 1: and that's where we might worry about inequality. Josh, talk

237
00:14:52,600 --> 00:14:55,080
Speaker 1: to us a little bit also about this idea explored

238
00:14:55,120 --> 00:14:59,680
Speaker 1: in your book on how artificial intelligence might actually contribute

239
00:14:59,680 --> 00:15:04,320
Speaker 1: to the concentration of certain industries in our economy, even

240
00:15:04,360 --> 00:15:09,240
Speaker 1: aid the creation of monopolies. Well, I think this is

241
00:15:09,280 --> 00:15:13,200
Speaker 1: something that has occurred with any digital platforms, especially ones

242
00:15:13,280 --> 00:15:17,160
Speaker 1: that work better with larger scale. So the current AI

243
00:15:17,360 --> 00:15:20,320
Speaker 1: runs on machine learning, and I have to emphasize the

244
00:15:20,400 --> 00:15:25,680
Speaker 1: learning part. You get better by putting your machines out

245
00:15:25,680 --> 00:15:30,840
Speaker 1: in the field and continuously adjusting them with new data

246
00:15:30,920 --> 00:15:34,080
Speaker 1: and new learning. In so, what that means is that

247
00:15:34,120 --> 00:15:37,040
Speaker 1: a company such as Google that has a lot more

248
00:15:37,560 --> 00:15:40,520
Speaker 1: people using its search engine is able to use AI

249
00:15:40,640 --> 00:15:43,920
Speaker 1: to improve that those search engines at a greater rate

250
00:15:43,960 --> 00:15:47,720
Speaker 1: than competitors such as Being and Duct Duct Go, and

251
00:15:47,960 --> 00:15:52,240
Speaker 1: so in that regard, you could have the preservation of

252
00:15:52,400 --> 00:15:55,840
Speaker 1: a dominance or the emergence of dominance. And similarly this

253
00:15:55,920 --> 00:15:59,040
Speaker 1: might hold for companies like Facebook, and it might hold

254
00:15:59,080 --> 00:16:02,760
Speaker 1: to a degree for companies like Apple and Amazon as well.

255
00:16:03,120 --> 00:16:05,680
Speaker 1: They just have a lot more activity and so their

256
00:16:05,720 --> 00:16:09,480
Speaker 1: potential for to to use AI to learn at a

257
00:16:09,560 --> 00:16:13,480
Speaker 1: faster rate might be there. But as with all of

258
00:16:13,520 --> 00:16:16,600
Speaker 1: these things, is sometimes these firms do get into a rut,

259
00:16:16,680 --> 00:16:20,680
Speaker 1: and sometimes people find better ways of learning and doing

260
00:16:20,720 --> 00:16:24,320
Speaker 1: things that might initially perform worse but have a better trajectory.

261
00:16:24,880 --> 00:16:26,600
Speaker 1: So it's not a given that we're going to have

262
00:16:27,120 --> 00:16:30,760
Speaker 1: the current monopolies, be the future monopolies or anything like that,

263
00:16:31,040 --> 00:16:34,000
Speaker 1: or the current large companies, but we might see new

264
00:16:34,040 --> 00:16:37,160
Speaker 1: ones develop on the basis of new tools. Every other time,

265
00:16:37,200 --> 00:16:40,800
Speaker 1: we've had a large technological revolution that has occurred, I'd

266
00:16:40,800 --> 00:16:44,240
Speaker 1: expect it to occur this time too. Just taking a

267
00:16:44,400 --> 00:16:50,040
Speaker 1: broader view of competition, you also get into which country

268
00:16:50,160 --> 00:16:52,800
Speaker 1: might be the dominant force in AI. If there's a

269
00:16:52,840 --> 00:16:57,080
Speaker 1: dominant force, the US has already has a clear lead,

270
00:16:57,480 --> 00:17:01,920
Speaker 1: and yet there's a lot of activity going on in China.

271
00:17:02,000 --> 00:17:06,000
Speaker 1: How do you see China ascending and competing against the

272
00:17:06,080 --> 00:17:10,200
Speaker 1: United States in AI. Well, this is where being able

273
00:17:10,240 --> 00:17:12,680
Speaker 1: to have access to the right sort of data comes

274
00:17:12,760 --> 00:17:15,880
Speaker 1: to play, not just data but also talent. So let's

275
00:17:15,880 --> 00:17:18,800
Speaker 1: talk about data first. The United States is sort of

276
00:17:18,800 --> 00:17:22,399
Speaker 1: a middle ground in terms of privacy regulation. Europe is

277
00:17:22,560 --> 00:17:26,040
Speaker 1: far more stringent, but China there's none at all, and

278
00:17:26,119 --> 00:17:29,920
Speaker 1: so Chinese firms have the ability to appropriate and use

279
00:17:30,359 --> 00:17:33,920
Speaker 1: consumer data to develop AI at a much greater rate

280
00:17:34,240 --> 00:17:37,160
Speaker 1: than you would be able to do inside the United States.

281
00:17:37,400 --> 00:17:41,280
Speaker 1: So there's a benefit. Secondly, there's the issue of capabilities.

282
00:17:41,800 --> 00:17:44,600
Speaker 1: At the moment, AI resources are thin on the ground.

283
00:17:44,720 --> 00:17:47,959
Speaker 1: It's hard to hire engineers they command six or seven

284
00:17:48,000 --> 00:17:51,080
Speaker 1: figure sums. But the United States at the moment is

285
00:17:51,080 --> 00:17:54,639
Speaker 1: cutting itself off from the global pool of talent in

286
00:17:55,600 --> 00:17:58,640
Speaker 1: machine learning, and this is something other countries aren't doing.

287
00:17:58,680 --> 00:18:01,639
Speaker 1: Not only they not cutting them selves off, They're also

288
00:18:01,760 --> 00:18:05,480
Speaker 1: providing resources. China is spending several billion dollars on a

289
00:18:05,520 --> 00:18:09,280
Speaker 1: technology cluster in this area. Russia are doing the same,

290
00:18:10,080 --> 00:18:14,199
Speaker 1: and even countries like Canada, the government's actively supporting the

291
00:18:14,240 --> 00:18:18,320
Speaker 1: development of news superclusters in the space. That's going to

292
00:18:18,400 --> 00:18:22,439
Speaker 1: help attract talent around the world, because talent wants to

293
00:18:22,440 --> 00:18:25,560
Speaker 1: be able to work, and I think that's another area

294
00:18:25,560 --> 00:18:29,040
Speaker 1: where the United States faces some risks. Let's end this

295
00:18:29,119 --> 00:18:33,080
Speaker 1: interview with the existential question that you address in your

296
00:18:33,080 --> 00:18:35,520
Speaker 1: book or try to address. We've all seen our share

297
00:18:35,560 --> 00:18:39,000
Speaker 1: of science fiction movies. The Rise of the Robots Terminator

298
00:18:39,080 --> 00:18:42,879
Speaker 1: to looms very large in my mind for me, josh

299
00:18:43,200 --> 00:18:45,640
Speaker 1: is this the end of the world as we know it? Well,

300
00:18:45,680 --> 00:18:49,000
Speaker 1: as we say in the book, not enough time yet

301
00:18:49,040 --> 00:18:51,280
Speaker 1: to tell. But you've got enough time to read our

302
00:18:51,320 --> 00:18:54,199
Speaker 1: book and be on the right side of it. All right, Well,

303
00:18:54,240 --> 00:18:57,240
Speaker 1: let's send it there. Joshua Against from the University of Toronto,

304
00:18:57,359 --> 00:19:00,280
Speaker 1: author of Prediction Machines, thank you very much for running

305
00:19:00,320 --> 00:19:06,800
Speaker 1: us on Benchmark. Thank you. Benchmark will be back next week.

306
00:19:06,920 --> 00:19:09,240
Speaker 1: Until then, you can find us on the Bloomberg terminal

307
00:19:09,240 --> 00:19:13,360
Speaker 1: Bloomberg dot com. Our Bloomberg app and podcast destinations such

308
00:19:13,400 --> 00:19:17,640
Speaker 1: as Apple Podcast, Spotify, or wherever you listen. We'd love

309
00:19:17,680 --> 00:19:19,679
Speaker 1: it if you took the time to post a review

310
00:19:19,720 --> 00:19:22,439
Speaker 1: of the show so more listeners can find us. You

311
00:19:22,480 --> 00:19:25,200
Speaker 1: can also check us out on Twitter, follow me at

312
00:19:25,280 --> 00:19:30,000
Speaker 1: at scott Landman, Chris You're at Chris Ja Conden, and

313
00:19:30,160 --> 00:19:33,240
Speaker 1: our guest Josh Gannes is at Josh gannes g. A

314
00:19:33,520 --> 00:19:37,080
Speaker 1: n S Benchmark is produced by Toper Foreheaz. The head

315
00:19:37,119 --> 00:19:40,720
Speaker 1: of Bloomberg Podcasts is Francesca Levy. Thanks for listening. To

316
00:19:40,800 --> 00:19:41,600
Speaker 1: see you next time.