1
00:00:03,840 --> 00:00:08,840
Speaker 1: This is Bloomberg Surveillance. The UK needs its own sovereignty,

2
00:00:08,920 --> 00:00:11,880
Speaker 1: it doesn't need rules set out for it by the

3
00:00:11,960 --> 00:00:16,040
Speaker 1: European community. That decisions to raise rates are not being

4
00:00:16,120 --> 00:00:19,000
Speaker 1: driven by inflation, so what is driving at There is

5
00:00:19,040 --> 00:00:21,599
Speaker 1: still only one global oil market and the price goes

6
00:00:21,680 --> 00:00:24,439
Speaker 1: up because of outages in Nigeria, and whether we're importing

7
00:00:24,440 --> 00:00:27,360
Speaker 1: oils from Nigeria not to reporting to be reflected here

8
00:00:27,360 --> 00:00:30,960
Speaker 1: at Bloomberg Surveillance, your link to the world of economics,

9
00:00:31,160 --> 00:00:35,120
Speaker 1: finance and investment on Bloomberg Radio. Good morning everyone, Michael

10
00:00:35,200 --> 00:00:37,720
Speaker 1: McKee and Tom Kane. The drag press conference will see

11
00:00:37,720 --> 00:00:43,000
Speaker 1: that in thirty minutes, low expectations, the official surveillance radar

12
00:00:43,120 --> 00:00:46,720
Speaker 1: is up for what Mr drog may say, not only

13
00:00:46,720 --> 00:00:49,040
Speaker 1: in the first five ten minutes of the press conference,

14
00:00:49,520 --> 00:00:53,159
Speaker 1: usually obligatory a yawn, But then it always seems to

15
00:00:53,200 --> 00:00:56,280
Speaker 1: get interesting. The Euro one eleven eighties seven, I'm gonna

16
00:00:56,280 --> 00:00:59,800
Speaker 1: call it weaker euro over the last few days worldwide,

17
00:01:00,240 --> 00:01:03,840
Speaker 1: across the nation. In New York, Bloomberg Surveillance vatch Vy

18
00:01:03,920 --> 00:01:08,119
Speaker 1: Cone Resnick, look Ahead, Gain insight, imagine more, get forward

19
00:01:08,160 --> 00:01:12,880
Speaker 1: thinking advice that can help turn business possibilities in the

20
00:01:13,000 --> 00:01:17,120
Speaker 1: business opportunities. Find out more at Cone Resnick dot com.

21
00:01:17,280 --> 00:01:19,959
Speaker 1: This is gonna be interesting, Michael McKee on the Bloomberg

22
00:01:19,959 --> 00:01:23,800
Speaker 1: There is t y L go, that Taylor rule go,

23
00:01:24,560 --> 00:01:27,640
Speaker 1: and you can plug in and hug the Newtonian mechanics

24
00:01:28,400 --> 00:01:32,759
Speaker 1: of the algebraic function, which approximates John B. Taylor's great

25
00:01:32,760 --> 00:01:36,399
Speaker 1: work out at Stanford. But there's other ways to do

26
00:01:36,480 --> 00:01:40,440
Speaker 1: monetary policy, aren't there approximates? Of course John always said

27
00:01:40,440 --> 00:01:43,040
Speaker 1: it wasn't a way to forecast where rates should be,

28
00:01:43,120 --> 00:01:45,919
Speaker 1: just a way to double check rate where rates should

29
00:01:45,920 --> 00:01:48,800
Speaker 1: have been in the past. But the question is can

30
00:01:48,880 --> 00:01:51,480
Speaker 1: you come up with rules? Can you come up with

31
00:01:52,120 --> 00:01:58,320
Speaker 1: some sort of reliable way of setting rates based on

32
00:01:58,480 --> 00:02:01,160
Speaker 1: data that takes human emotion sort of out of it

33
00:02:01,720 --> 00:02:04,640
Speaker 1: um And that's been a focus of the work of

34
00:02:04,640 --> 00:02:07,920
Speaker 1: Michael Currn's. He's a professor and a National Center Chair

35
00:02:07,960 --> 00:02:10,880
Speaker 1: at the University of Pennsylvania down in Philadelphia, where, of

36
00:02:10,960 --> 00:02:14,760
Speaker 1: course Janet yellen Um, the real Janet you know, not

37
00:02:14,800 --> 00:02:18,040
Speaker 1: the computerized version, is speaking on Monday, and that highly

38
00:02:18,080 --> 00:02:23,080
Speaker 1: anticipated speech are Christopher Connon talked with you, professor, not

39
00:02:23,160 --> 00:02:27,400
Speaker 1: long ago, about the possibility of using artificial intelligence to

40
00:02:28,120 --> 00:02:32,280
Speaker 1: improve economic forecasting. And I guess that's the key, uh,

41
00:02:32,400 --> 00:02:34,760
Speaker 1: to to improve forecasting, and then you can use a

42
00:02:34,840 --> 00:02:37,520
Speaker 1: rule based on the forecast to figure out where your

43
00:02:37,560 --> 00:02:41,760
Speaker 1: rate should be. Yeah, that's right. Chris and I had

44
00:02:41,760 --> 00:02:45,519
Speaker 1: a long conversation about the possible application of machine learning

45
00:02:45,560 --> 00:02:49,160
Speaker 1: to sort of macro economic forecasting in general and kind

46
00:02:49,160 --> 00:02:53,040
Speaker 1: of policy setting more specifically, and UM, you know what

47
00:02:53,080 --> 00:02:56,960
Speaker 1: I told Chris was that, UM, I'm sort of a

48
00:02:57,040 --> 00:02:59,200
Speaker 1: machine learning advocate in the sense that I think that

49
00:02:59,400 --> 00:03:02,720
Speaker 1: such a thing as possible in principle, but probably pretty

50
00:03:02,760 --> 00:03:06,960
Speaker 1: far from being practical at this point. Yeah. I mean,

51
00:03:07,400 --> 00:03:12,359
Speaker 1: it's not like the FED and other commercial enterprises don't

52
00:03:12,400 --> 00:03:15,680
Speaker 1: have models, but they find that there are so many

53
00:03:15,760 --> 00:03:19,840
Speaker 1: millions of influences that it's almost impossible to find tune

54
00:03:19,840 --> 00:03:23,240
Speaker 1: them enough. Yeah, that's right. And I think that, you know,

55
00:03:23,320 --> 00:03:25,840
Speaker 1: with all of the kind of media frenzy about machine

56
00:03:25,919 --> 00:03:29,680
Speaker 1: learning these days, we forget that, um many people in

57
00:03:29,720 --> 00:03:32,720
Speaker 1: many fields have been doing what has now been rebranded

58
00:03:32,760 --> 00:03:35,960
Speaker 1: machine learning for a very long time. I think in

59
00:03:36,080 --> 00:03:39,760
Speaker 1: terms of kind of macroeconomic policy prediction, the thing that's

60
00:03:39,800 --> 00:03:44,240
Speaker 1: particularly difficult at this point for machines is just incorporating

61
00:03:44,480 --> 00:03:48,160
Speaker 1: knowledge about you know, how markets work, how policy interacts

62
00:03:48,160 --> 00:03:53,320
Speaker 1: with markets, how international events shape you know, the future. UM,

63
00:03:53,440 --> 00:03:57,840
Speaker 1: and it's just very difficult to get clean data that

64
00:03:58,280 --> 00:04:01,360
Speaker 1: you know kind of UM relates all of those working

65
00:04:01,440 --> 00:04:04,440
Speaker 1: parts over a long enough history that you would have

66
00:04:04,480 --> 00:04:07,600
Speaker 1: seen cycles, you would have seen market crashes, you would

67
00:04:07,600 --> 00:04:10,080
Speaker 1: have had a lot of examples of how economy has

68
00:04:10,160 --> 00:04:14,120
Speaker 1: changed and responses to policy. That goes to the heart

69
00:04:14,120 --> 00:04:17,040
Speaker 1: of the matter. And course this goes to artificial intelligence

70
00:04:17,080 --> 00:04:20,719
Speaker 1: and the rest. Where what have we accomplished in the

71
00:04:20,800 --> 00:04:24,320
Speaker 1: last thirty years? If Governor Tarula who was just on

72
00:04:24,440 --> 00:04:28,240
Speaker 1: with us, Cherry Yellen, Mario Doroggi, who's gonna speak in

73
00:04:28,279 --> 00:04:32,280
Speaker 1: twenty five minutes. If people like that are working with

74
00:04:32,360 --> 00:04:38,440
Speaker 1: basic algebra, Newtonian mechanics, and maybe something is fungible, is dynamics,

75
00:04:38,520 --> 00:04:44,760
Speaker 1: stochastic general equilibrium theory, How can AI help them? I mean,

76
00:04:44,800 --> 00:04:48,520
Speaker 1: it's very to me. It's your world is very nonlinear,

77
00:04:49,240 --> 00:04:53,160
Speaker 1: it's it's got huge degrees of freedom issues. How can

78
00:04:53,360 --> 00:04:58,400
Speaker 1: AI assist economists to do a better job? Well, I

79
00:04:58,440 --> 00:05:02,159
Speaker 1: mean I think A can help, and in machine learning

80
00:05:02,160 --> 00:05:06,200
Speaker 1: more specifically, can help in any domain in which you

81
00:05:06,279 --> 00:05:10,320
Speaker 1: have massive amounts of historical data, including very high dimensional

82
00:05:10,400 --> 00:05:14,960
Speaker 1: data um that's sort of relatively clean and and sort

83
00:05:15,000 --> 00:05:18,000
Speaker 1: of drawn under you know, kind of similar conditions over

84
00:05:18,040 --> 00:05:21,040
Speaker 1: a long period of time. The more sort of you know,

85
00:05:21,520 --> 00:05:25,480
Speaker 1: heterogeneous your data is, the more diversity is, the harder

86
00:05:25,560 --> 00:05:27,640
Speaker 1: it is to kind of apply machine learning. And this

87
00:05:27,680 --> 00:05:31,799
Speaker 1: is why you see machine learning especially succeeding in either

88
00:05:31,880 --> 00:05:35,160
Speaker 1: cases where there's a massive amount of clean data, as

89
00:05:35,160 --> 00:05:39,360
Speaker 1: in areas like speech recognition or exactly, or in sort

90
00:05:39,360 --> 00:05:42,839
Speaker 1: of very closed environments like game playing like chests or go,

91
00:05:43,520 --> 00:05:47,200
Speaker 1: which is difficult as those problems are. You know you're

92
00:05:47,200 --> 00:05:49,440
Speaker 1: in a closed world, right, you know that the rules

93
00:05:49,480 --> 00:05:52,320
Speaker 1: are very Yeah, well there's I get the beautifully explained.

94
00:05:52,320 --> 00:05:55,680
Speaker 1: There's sixty four squares and chess, I get it. Can

95
00:05:55,760 --> 00:06:00,359
Speaker 1: you take your world and help Janet Yellen with the

96
00:06:00,480 --> 00:06:03,680
Speaker 1: data that goes back to Lawrence client at pen In

97
00:06:05,000 --> 00:06:08,080
Speaker 1: or whatever. I mean, was that data then clean or

98
00:06:08,160 --> 00:06:11,679
Speaker 1: is that data now clean? I think in the case

99
00:06:11,800 --> 00:06:14,800
Speaker 1: I mean, and and mind you, I'm not an economist

100
00:06:14,800 --> 00:06:17,479
Speaker 1: per se, you know, and even less we take that

101
00:06:17,600 --> 00:06:23,240
Speaker 1: as an advantage. Certain I appreciate that, but I think

102
00:06:23,279 --> 00:06:26,360
Speaker 1: my short answer to that is no, and it's not

103
00:06:26,480 --> 00:06:28,320
Speaker 1: so much whether the data is clean or not, it's

104
00:06:28,360 --> 00:06:30,240
Speaker 1: just sort of the length of history. I mean to

105
00:06:30,560 --> 00:06:33,599
Speaker 1: give an analogy, right, Um, you know, if you go

106
00:06:33,720 --> 00:06:36,680
Speaker 1: back thirty years before sort of the automation of Wall

107
00:06:36,720 --> 00:06:39,440
Speaker 1: Street more generally, there were a lot of you know,

108
00:06:39,560 --> 00:06:44,440
Speaker 1: trading problems that at that time really required human expertise, intuition,

109
00:06:44,560 --> 00:06:47,080
Speaker 1: knowledge of how markets work, how they might react to

110
00:06:47,200 --> 00:06:50,400
Speaker 1: certain exogenous events, and so on and so forth. And

111
00:06:50,440 --> 00:06:52,280
Speaker 1: now a lot of that stuff has been automated, right,

112
00:06:52,279 --> 00:06:54,320
Speaker 1: I mean, sort of the brokerage business has become a

113
00:06:54,400 --> 00:06:57,440
Speaker 1: very difficult business, um, at least being a sort of

114
00:06:57,440 --> 00:07:00,599
Speaker 1: the human sort of specialist brokerage. But this has become

115
00:07:00,680 --> 00:07:04,279
Speaker 1: very difficult as algorithmic trading has risen. And the real

116
00:07:04,320 --> 00:07:07,400
Speaker 1: reason now that the trading has taken over is partly

117
00:07:07,400 --> 00:07:11,720
Speaker 1: the automation. But the automation has you know, generated years

118
00:07:11,720 --> 00:07:16,240
Speaker 1: and years a very clean data. Um. That makes machine

119
00:07:16,320 --> 00:07:19,560
Speaker 1: learning a much more practical, you know, sort of practical approach.

120
00:07:19,880 --> 00:07:22,440
Speaker 1: And by the way, you tend to see the you know,

121
00:07:22,520 --> 00:07:26,960
Speaker 1: the greatest uses of machine learning in trading or economic

122
00:07:27,040 --> 00:07:31,040
Speaker 1: settings where the most data is generated, so high frequency

123
00:07:31,120 --> 00:07:33,760
Speaker 1: trading for instance, right, just because of the speed at

124
00:07:33,800 --> 00:07:36,840
Speaker 1: which thing is things are happening, you know, in the

125
00:07:36,920 --> 00:07:40,000
Speaker 1: same minute. Um, I get way more data about high

126
00:07:40,040 --> 00:07:43,200
Speaker 1: frequency trading that I do about macro economous policy predictions.

127
00:07:43,240 --> 00:07:45,000
Speaker 1: So can your model tell me if there's going to

128
00:07:45,080 --> 00:07:52,400
Speaker 1: be a junior July rate increase? I couldn't tell you

129
00:07:52,440 --> 00:07:55,240
Speaker 1: if I knew good. No one else can either, So

130
00:07:55,360 --> 00:07:59,680
Speaker 1: can like, like one more question? Please? Well, I just

131
00:07:59,720 --> 00:08:02,680
Speaker 1: car is because Bill Dudley and others are de fitive

132
00:08:02,720 --> 00:08:06,560
Speaker 1: noted that one failure of their models is that doesn't

133
00:08:06,600 --> 00:08:10,880
Speaker 1: take financial markets into account. Financial markets, much of it

134
00:08:10,920 --> 00:08:13,000
Speaker 1: is high frequency trading, but much of it is also

135
00:08:13,200 --> 00:08:19,520
Speaker 1: market psychology. How does machine learning account for psychological factors? Yeah,

136
00:08:19,560 --> 00:08:22,040
Speaker 1: so that's a great question. And um, one of you

137
00:08:22,080 --> 00:08:25,440
Speaker 1: mentioned equilibrium a little while ago, and you know, the

138
00:08:25,560 --> 00:08:30,200
Speaker 1: term equilibrium kind of refers to taking the strategic considerations

139
00:08:30,360 --> 00:08:33,480
Speaker 1: of the parties into account, you know, or or you know,

140
00:08:33,559 --> 00:08:35,800
Speaker 1: kind of counter factuals, if you like, sort of not

141
00:08:35,960 --> 00:08:39,480
Speaker 1: just how did the data look historically, but how might

142
00:08:39,559 --> 00:08:42,360
Speaker 1: it have looked different if we had done something differently

143
00:08:42,400 --> 00:08:45,840
Speaker 1: and something different had happened. And one thing I like to,

144
00:08:46,200 --> 00:08:47,520
Speaker 1: you know, one way I like to put it when

145
00:08:47,520 --> 00:08:51,120
Speaker 1: I talk to people about machine learning in finance or

146
00:08:51,120 --> 00:08:54,439
Speaker 1: economics versus in other domains, which is, you know, as

147
00:08:54,520 --> 00:08:57,840
Speaker 1: hard as the problem of let's say, recognizing whether there's

148
00:08:57,840 --> 00:09:00,480
Speaker 1: a cat in a video on YouTube or not might be,

149
00:09:00,720 --> 00:09:03,640
Speaker 1: and you know, you might think I'm joking, but actually

150
00:09:03,640 --> 00:09:07,000
Speaker 1: that's not an easy problem. As hard as that problem

151
00:09:07,080 --> 00:09:10,000
Speaker 1: might be, it one one advantage you have an applying

152
00:09:10,000 --> 00:09:12,319
Speaker 1: machine learning to such a problem is that you can

153
00:09:12,320 --> 00:09:15,959
Speaker 1: be pretty sure that you know the world is not

154
00:09:16,720 --> 00:09:19,920
Speaker 1: you know that your very effort to decide whether there's

155
00:09:19,960 --> 00:09:22,520
Speaker 1: a cat in the video or not is not you know,

156
00:09:22,640 --> 00:09:26,400
Speaker 1: changing the nature and casts um you know themselves right.

157
00:09:26,880 --> 00:09:28,960
Speaker 1: Where this is not true in trading, you can be

158
00:09:29,040 --> 00:09:32,440
Speaker 1: quite sure that you know your very effort to predict

159
00:09:32,559 --> 00:09:36,400
Speaker 1: something in financial markets and then act on that prediction

160
00:09:37,120 --> 00:09:39,880
Speaker 1: link fact change the markets in a way that makes

161
00:09:40,160 --> 00:09:43,240
Speaker 1: what you're doing less effective. Right, So it's that kind

162
00:09:43,240 --> 00:09:46,839
Speaker 1: of adaptive dynamic, the market reacting to what you're doing

163
00:09:47,120 --> 00:09:51,320
Speaker 1: because of the strategic considerations the other parties involved, and

164
00:09:51,360 --> 00:09:54,280
Speaker 1: there there are branches of machine learning that are trying

165
00:09:54,280 --> 00:09:59,280
Speaker 1: to seriously take those kind of strategic considerations and counter

166
00:09:59,320 --> 00:10:02,120
Speaker 1: factuals in to account. But it's a much much more

167
00:10:02,120 --> 00:10:05,360
Speaker 1: difficult problem. And I think that you know, we're very

168
00:10:05,440 --> 00:10:08,080
Speaker 1: very far from kind of understanding how to deal with

169
00:10:08,120 --> 00:10:10,240
Speaker 1: such problems at the at the kind of scale that

170
00:10:10,280 --> 00:10:12,520
Speaker 1: you guys are talking about. Michael Currents, thank you so much.

171
00:10:12,559 --> 00:10:14,839
Speaker 1: At the University of Pennsylvania and machine learning and a

172
00:10:14,920 --> 00:10:18,760
Speaker 1: touch there and AI as well, our machine learning, I mean,

173
00:10:18,760 --> 00:10:22,600
Speaker 1: we are robots, Mike Is. We're twenty minutes away from

174
00:10:22,600 --> 00:10:25,360
Speaker 1: the drug press comments. I would don't, Mike. Not much

175
00:10:25,400 --> 00:10:27,960
Speaker 1: movement in the market. I guess that's not a surprise, Mike.

176
00:10:28,000 --> 00:10:29,719
Speaker 1: I would know two days in a row, strong and

177
00:10:29,760 --> 00:10:33,280
Speaker 1: strong Yen one O eight ninety two and a yen

178
00:10:33,360 --> 00:10:37,080
Speaker 1: as well, indeed, but but no real reaction to Torulo

179
00:10:37,600 --> 00:10:39,720
Speaker 1: or to the e c B. And I guess, uh

180
00:10:39,960 --> 00:10:42,520
Speaker 1: Jim vot will put it, well, it's all down to

181
00:10:42,600 --> 00:10:45,880
Speaker 1: Janet Yellen on Monday, the futures and negative four down

182
00:10:45,880 --> 00:10:52,920
Speaker 1: futures at negative yield one point eight three. Time now

183
00:10:52,960 --> 00:10:54,640
Speaker 1: to check in with Michael Barr and get caught up

184
00:10:54,640 --> 00:10:57,040
Speaker 1: on world in the national headlines. Mike, Tom, thank you

185
00:10:57,160 --> 00:10:59,920
Speaker 1: very much. Hillary Clinton has sat to unleash a four

186
00:11:00,000 --> 00:11:04,040
Speaker 1: in policy attack on Donald Trump. The former Secretary of

187
00:11:04,040 --> 00:11:06,360
Speaker 1: State will use a speech in San Diego today to

188
00:11:06,400 --> 00:11:11,080
Speaker 1: cast the Republican as unqualified and dangerous. Trump accused Clinton

189
00:11:11,120 --> 00:11:13,839
Speaker 1: of lying about his foreign policy plans at a rally

190
00:11:14,120 --> 00:11:18,240
Speaker 1: in Sacramento, California, last night. Classes will resume next week.

191
00:11:18,320 --> 00:11:21,360
Speaker 1: At u c l A yesterday, a professor was shot

192
00:11:21,360 --> 00:11:24,439
Speaker 1: and killed before police say the gunman took his own life.

193
00:11:24,920 --> 00:11:28,120
Speaker 1: The head of the Global Airline Industry Association says terrorist

194
00:11:28,200 --> 00:11:32,640
Speaker 1: attacks will not stop surging travel demand. International air Transport

195
00:11:32,640 --> 00:11:37,160
Speaker 1: Association CEO Tony Tyler told Bloomberg people don't get frightened

196
00:11:37,160 --> 00:11:40,480
Speaker 1: off by these thugs. Global News twenty four hours a day,

197
00:11:40,520 --> 00:11:43,480
Speaker 1: powered by our twenty four hundred journalists more than a

198
00:11:43,559 --> 00:11:46,360
Speaker 1: hundred fifty news bureaus around the world. I'm Michael Barr.

199
00:11:46,960 --> 00:11:49,240
Speaker 1: Michael Barr, Thanks so much, che life. If you've got

200
00:11:49,240 --> 00:11:51,880
Speaker 1: a Bloomberg terminal in your car, t l i V

201
00:11:52,920 --> 00:11:56,240
Speaker 1: great updates on the Draggy press conference and on OPEC

202
00:11:56,880 --> 00:11:59,600
Speaker 1: in vi Enna. Futures at negative four. This is Bloomberg.

203
00:11:59,640 --> 00:12:07,840
Speaker 1: Save us. The news update brought you by your Mercedes

204
00:12:07,840 --> 00:12:09,920
Speaker 1: Benz tri state dealer. The star you've been wishing for

205
00:12:10,000 --> 00:12:12,000
Speaker 1: is waiting at the Mercedes Benz Summer Event. Head to

206
00:12:12,000 --> 00:12:14,600
Speaker 1: your Mercedes Benz Tri State dealer for more details. Hurry

207
00:12:14,840 --> 00:12:19,240
Speaker 1: before this star flies by.