1
00:00:05,900 --> 00:00:09,760
Speaker 1: Hello, this is Copy Time, a podcast series on Markets

2
00:00:09,770 --> 00:00:13,309
Speaker 1: and Economies from DVS Group Research. I'm Tamur Bank, chief economist.

3
00:00:13,319 --> 00:00:16,280
Speaker 1: Welcoming you to our 126th episode

4
00:00:16,930 --> 00:00:20,069
Speaker 1: today. It's all about A I and to be more

5
00:00:20,079 --> 00:00:23,270
Speaker 1: specific about generative A I, we will hear about the

6
00:00:23,280 --> 00:00:27,170
Speaker 1: subject of the moment from Microsoft's Asia head of business development.

7
00:00:27,729 --> 00:00:32,139
Speaker 1: Zia Zaman. Zia has had decades of experience on growth

8
00:00:32,150 --> 00:00:36,348
Speaker 1: and innovation, both at Microsoft and previously at metlife Lumen

9
00:00:36,360 --> 00:00:40,189
Speaker 1: Lab and a whole bunch of other companies here in Asia,

10
00:00:40,810 --> 00:00:42,310
Speaker 1: Zia Zaman. Welcome to Kobe Time.

11
00:00:43,110 --> 00:00:44,930
Speaker 2: Wonderful to be here, Tamar. Thank you so much.

12
00:00:45,080 --> 00:00:48,069
Speaker 1: It's great to have you Xia. Um I'm gonna start

13
00:00:48,080 --> 00:00:52,099
Speaker 1: out a bit provocatively. Uh The iphone was rolled out

14
00:00:52,110 --> 00:00:55,860
Speaker 1: in mid 2007 and I recall distinctly that by the

15
00:00:55,869 --> 00:00:59,770
Speaker 1: end of 2008, from the hardware production ecosystem to the

16
00:00:59,779 --> 00:01:03,250
Speaker 1: smartphone app development business. Not to mention the way we

17
00:01:03,259 --> 00:01:07,279
Speaker 1: communicated fundamentally, there was a full blown mobile revolution in

18
00:01:07,290 --> 00:01:07,729
Speaker 1: the world.

19
00:01:08,330 --> 00:01:10,860
Speaker 1: Now, it's been about the same span of time since

20
00:01:10,870 --> 00:01:15,290
Speaker 1: Chad GP T came into our consciousness. Late 2023. Uh

21
00:01:15,300 --> 00:01:18,988
Speaker 1: has the world been transformed in a comparable manner.

22
00:01:21,180 --> 00:01:24,199
Speaker 2: I think it really has Timor and thank you for asking.

23
00:01:24,209 --> 00:01:29,129
Speaker 2: I think the iphone moment uh was for me, the

24
00:01:29,160 --> 00:01:34,029
Speaker 2: uh biggest and maybe the only major disruption we had

25
00:01:34,040 --> 00:01:38,550
Speaker 2: in the 2000 to 2015 time frame. I actually think

26
00:01:38,559 --> 00:01:41,889
Speaker 2: that those, that decade, the the the Knots was probably

27
00:01:41,940 --> 00:01:45,959
Speaker 2: um one of the least productive times for tech.

28
00:01:46,180 --> 00:01:50,419
Speaker 2: Um we had tremendous growth in tech um as a

29
00:01:50,430 --> 00:01:54,480
Speaker 2: result of the transistor of mainframes of client server of

30
00:01:54,489 --> 00:01:58,720
Speaker 2: the PC revolution. And then we hit the doldrums and

31
00:01:58,730 --> 00:02:03,199
Speaker 2: the single bright light was mo the mobile revolution. And

32
00:02:03,459 --> 00:02:07,319
Speaker 2: in many ways, I would liken what we're seeing with

33
00:02:07,330 --> 00:02:09,360
Speaker 2: generative A I and we know A I has been

34
00:02:09,369 --> 00:02:12,240
Speaker 2: around for 50 or 60 years, but the generative A

35
00:02:12,250 --> 00:02:12,520
Speaker 2: I

36
00:02:12,910 --> 00:02:17,919
Speaker 2: to be even greater in terms of the change that

37
00:02:17,929 --> 00:02:21,850
Speaker 2: it might have on work on the way in which

38
00:02:21,860 --> 00:02:23,959
Speaker 2: we live and the way which we commute, it's still

39
00:02:23,970 --> 00:02:27,000
Speaker 2: a tool, but I liken it perhaps more to the

40
00:02:27,008 --> 00:02:31,800
Speaker 2: 1994 Netscape Moment. Um Or even I see the parallels

41
00:02:31,809 --> 00:02:35,240
Speaker 2: back to the printing press in terms of it, enabling

42
00:02:35,369 --> 00:02:38,279
Speaker 2: a different set of language, a different way of thinking

43
00:02:38,288 --> 00:02:38,799
Speaker 2: about

44
00:02:39,399 --> 00:02:42,509
Speaker 2: um how we do things in a different type of

45
00:02:42,660 --> 00:02:44,100
Speaker 2: if you will, intelligence.

46
00:02:45,300 --> 00:02:50,929
Speaker 1: Are we doing things fundamentally differently already? Or you are

47
00:02:50,940 --> 00:02:52,800
Speaker 1: talking in aspirational terms?

48
00:02:54,050 --> 00:02:57,000
Speaker 2: I think that there are so many examples of companies

49
00:02:57,008 --> 00:03:01,309
Speaker 2: that are doing things differently already. And yet we are

50
00:03:01,320 --> 00:03:05,869
Speaker 2: still so early in this journey. If there's one thing

51
00:03:05,880 --> 00:03:08,609
Speaker 2: that I would like your listeners to think about is

52
00:03:08,619 --> 00:03:13,949
Speaker 2: that we haven't seen anything yet. There is a rule

53
00:03:13,960 --> 00:03:17,169
Speaker 2: which is sometimes attributed to Bill Gates,

54
00:03:17,440 --> 00:03:20,038
Speaker 2: which he actually didn't say it's actually the Amara rule,

55
00:03:20,050 --> 00:03:22,750
Speaker 2: which is that we overestimate the amount of change that

56
00:03:22,758 --> 00:03:25,020
Speaker 2: will happen in one or two years and underestimate the

57
00:03:25,029 --> 00:03:27,259
Speaker 2: amount of change that will happen over 10 years.

58
00:03:27,910 --> 00:03:30,329
Speaker 2: And when you think about this, this comes from the

59
00:03:30,339 --> 00:03:33,690
Speaker 2: institutes for the future, it has proven to be true

60
00:03:33,699 --> 00:03:38,169
Speaker 2: over and over again, we are in year two of 10.

61
00:03:39,039 --> 00:03:42,740
Speaker 2: And in that principle, the year one and year two

62
00:03:42,750 --> 00:03:47,759
Speaker 2: change is sometimes overblown. I worked for Gartner for three years.

63
00:03:47,830 --> 00:03:50,559
Speaker 2: There is a hype cycle in terms of the expectations.

64
00:03:50,759 --> 00:03:54,380
Speaker 2: But what happens as any technology matures is that you

65
00:03:54,389 --> 00:03:56,380
Speaker 2: start to see the benefits

66
00:03:56,460 --> 00:03:59,880
Speaker 2: through the second half of that decade if you will.

67
00:03:59,940 --> 00:04:02,059
Speaker 2: And in that year of 10 principle, I do want

68
00:04:02,070 --> 00:04:05,160
Speaker 2: to say say that we have, we haven't seen the

69
00:04:05,169 --> 00:04:09,000
Speaker 2: amount of change that we will yet. But you have

70
00:04:09,009 --> 00:04:11,899
Speaker 2: to remember that the 1st 1st iphone people said look

71
00:04:11,910 --> 00:04:13,759
Speaker 2: like an ipod with a screen

72
00:04:14,020 --> 00:04:17,380
Speaker 2: and it was only until the app store unlocked a

73
00:04:17,390 --> 00:04:21,640
Speaker 2: lot of value that we finally get the explosion, not

74
00:04:21,649 --> 00:04:24,159
Speaker 2: just of users but of different ways in which we

75
00:04:24,170 --> 00:04:26,890
Speaker 2: use it and the App Store itself was from N

76
00:04:26,899 --> 00:04:30,488
Speaker 2: TT Docomo. Um They were the first to establish something

77
00:04:30,500 --> 00:04:34,219
Speaker 2: like that. So it builds upon, build upon build

78
00:04:34,928 --> 00:04:39,040
Speaker 2: and eventually what we will get, um is um the

79
00:04:39,049 --> 00:04:40,670
Speaker 2: kind of change that I think we're talking about.

80
00:04:42,480 --> 00:04:45,820
Speaker 1: Um Yeah, I mean, I completely concur with you, but

81
00:04:45,829 --> 00:04:49,140
Speaker 1: I'm just going to play Devil's Africa for just a second.

82
00:04:49,470 --> 00:04:53,630
Speaker 1: Um which is, uh in the last few years, we've

83
00:04:53,640 --> 00:04:57,079
Speaker 1: had a few other hype cycles around and I think

84
00:04:57,089 --> 00:04:58,859
Speaker 1: it's kind of unfair to say their hype cycles because

85
00:04:58,869 --> 00:05:01,909
Speaker 1: I think there's something genuinely foundational about those developments, but

86
00:05:01,920 --> 00:05:04,058
Speaker 1: they just didn't become as big or ubiquitous. I've been

87
00:05:04,070 --> 00:05:08,339
Speaker 1: thinking about the metaverse or the Blockchain around cryptocurrencies and

88
00:05:08,350 --> 00:05:08,820
Speaker 1: so on.

89
00:05:09,089 --> 00:05:13,160
Speaker 1: Uh What gives you confidence that this wave of technology

90
00:05:13,170 --> 00:05:16,140
Speaker 1: adoption will be far bigger to your point that if

91
00:05:16,149 --> 00:05:17,988
Speaker 1: you haven't seen anything yet, it's going to be far

92
00:05:18,000 --> 00:05:18,869
Speaker 1: more consequential.

93
00:05:19,720 --> 00:05:22,308
Speaker 2: That's a great point. And uh many of us have

94
00:05:22,320 --> 00:05:27,190
Speaker 2: thought about the promise of um sort of decentralized computing and,

95
00:05:27,200 --> 00:05:29,260
Speaker 2: and I, and I've done a lot of work around

96
00:05:29,570 --> 00:05:33,510
Speaker 2: uh distributor ledger technology and what that might do and

97
00:05:33,519 --> 00:05:38,290
Speaker 2: yet the obstacles to roll out for that particular technology

98
00:05:38,380 --> 00:05:42,130
Speaker 2: um hit a wall and when it came to um

99
00:05:42,589 --> 00:05:47,040
Speaker 2: the metaverse, uh there is potential in the future as well.

100
00:05:47,089 --> 00:05:50,019
Speaker 2: But for now, I think that what we're seeing is

101
00:05:50,160 --> 00:05:53,929
Speaker 2: two steps forward, one step back, they are big unlocks

102
00:05:53,940 --> 00:05:57,049
Speaker 2: um as will be energy. I think that the big

103
00:05:57,059 --> 00:06:01,738
Speaker 2: unlocks around energy will open up fundamentally new economies

104
00:06:01,964 --> 00:06:05,284
Speaker 2: and opportunities for us. They may drive marginal costs down

105
00:06:05,295 --> 00:06:07,954
Speaker 2: to zero. But when you really think about how different

106
00:06:07,964 --> 00:06:10,834
Speaker 2: it is decentralized computing gave us a different model for

107
00:06:10,845 --> 00:06:14,404
Speaker 2: thinking about how we might organize. Um you know, Cryptocurrency

108
00:06:14,415 --> 00:06:17,243
Speaker 2: was kind of a bet against the US dollar in

109
00:06:17,255 --> 00:06:21,244
Speaker 2: many ways um as the the reserve currency of the world.

110
00:06:21,545 --> 00:06:24,535
Speaker 2: Um and the metaverse, um I'm a big fan of

111
00:06:24,545 --> 00:06:26,704
Speaker 2: Ready Player. One watched it, read it. And I do

112
00:06:26,714 --> 00:06:29,774
Speaker 2: believe that there's AAA cautionary tale in there,

113
00:06:30,100 --> 00:06:33,849
Speaker 2: but there's something dystopian about where some of those technologies

114
00:06:33,940 --> 00:06:36,729
Speaker 2: are leading us to. Whereas I fundamentally think of this

115
00:06:36,738 --> 00:06:41,109
Speaker 2: tool as maybe more utopian and unlocking more uh types

116
00:06:41,119 --> 00:06:43,890
Speaker 2: of value. And, and if you go to the printing press,

117
00:06:43,899 --> 00:06:46,959
Speaker 2: it unlocked a democratic way of thinking about the dissemination

118
00:06:46,970 --> 00:06:50,049
Speaker 2: of knowledge. It gave us tools, the internet that is

119
00:06:50,059 --> 00:06:53,799
Speaker 2: at the beginning, um that would allow more people to democratize.

120
00:06:53,809 --> 00:06:55,519
Speaker 2: Um If there's a second thing that I want the

121
00:06:55,529 --> 00:06:58,799
Speaker 2: people of this podcast to remember is how inclusive

122
00:06:59,290 --> 00:07:01,789
Speaker 2: some of the technologies are that we are going to

123
00:07:01,799 --> 00:07:03,820
Speaker 2: help develop. And I don't mean our company, I mean,

124
00:07:03,829 --> 00:07:06,570
Speaker 2: the people that are our partners that are doing great

125
00:07:06,579 --> 00:07:08,630
Speaker 2: work on top of the platform. And if I don't

126
00:07:08,640 --> 00:07:10,459
Speaker 2: give enough examples by the end of the hour around

127
00:07:10,470 --> 00:07:14,010
Speaker 2: the inclusive nature of tech today. Then uh then stop

128
00:07:14,019 --> 00:07:14,859
Speaker 2: me and I will try.

129
00:07:15,600 --> 00:07:19,149
Speaker 1: Uh yeah, you've touched on a bunch of threads. Uh Utopia, dystopia.

130
00:07:19,160 --> 00:07:21,290
Speaker 1: I certainly want to unpack with you later, but you

131
00:07:21,299 --> 00:07:24,859
Speaker 1: also talked about cost. So I want to delve into

132
00:07:24,869 --> 00:07:27,769
Speaker 1: that a little bit. The numbers that I see from

133
00:07:27,779 --> 00:07:31,040
Speaker 1: your company and a bunch of other tech leading companies.

134
00:07:31,049 --> 00:07:34,500
Speaker 1: It's just mind boggling tens of billions of dollars, not

135
00:07:34,510 --> 00:07:36,630
Speaker 1: just this year, the plan is to spend tens of

136
00:07:36,640 --> 00:07:38,899
Speaker 1: billions of dollars every year for the rest of the decade,

137
00:07:38,910 --> 00:07:42,820
Speaker 1: if not longer. So the investment going into operational and

138
00:07:42,829 --> 00:07:45,019
Speaker 1: commercializing large language model is just

139
00:07:45,549 --> 00:07:48,600
Speaker 1: astonishing. Why is it so costly? Give us a sense

140
00:07:48,609 --> 00:07:49,570
Speaker 1: of where the money is going.

141
00:07:50,570 --> 00:07:52,519
Speaker 2: Um Yeah, thanks for asking. I think one of the

142
00:07:52,529 --> 00:07:57,429
Speaker 2: things that we are recognizing is it is so prohibitively

143
00:07:57,440 --> 00:08:01,390
Speaker 2: costly to uh build and run these models that it,

144
00:08:01,399 --> 00:08:05,850
Speaker 2: it cannot be done at scale by hundreds or thousands

145
00:08:05,859 --> 00:08:06,750
Speaker 2: of people. It will

146
00:08:06,825 --> 00:08:10,644
Speaker 2: probably be a handful under 10 companies that will put

147
00:08:10,654 --> 00:08:13,143
Speaker 2: in the tens of billions of dollars in order to

148
00:08:13,154 --> 00:08:16,885
Speaker 2: build the infrastructure to train and deploy these models. And you,

149
00:08:16,894 --> 00:08:19,915
Speaker 2: you think about that as um just the, the minimum

150
00:08:19,924 --> 00:08:22,325
Speaker 2: efficient scale to, to, to run the model and that,

151
00:08:22,464 --> 00:08:22,954
Speaker 2: that

152
00:08:23,279 --> 00:08:26,579
Speaker 2: floor just keeps going up and up and up. And

153
00:08:26,589 --> 00:08:31,019
Speaker 2: it's partly as a result of the exponential uh benefit

154
00:08:31,029 --> 00:08:34,500
Speaker 2: that comes from scale when it comes to learning. Now,

155
00:08:34,510 --> 00:08:36,780
Speaker 2: let's remember what the large language model does is it

156
00:08:36,789 --> 00:08:40,409
Speaker 2: makes a prediction about the next best word that is statistical.

157
00:08:40,419 --> 00:08:44,340
Speaker 2: And when that statistical prediction gets better as a result

158
00:08:44,349 --> 00:08:49,739
Speaker 2: of having more parameters or inputs, it just simply means that,

159
00:08:49,750 --> 00:08:50,270
Speaker 2: that

160
00:08:50,580 --> 00:08:54,459
Speaker 2: N plus one prediction is so much better when N

161
00:08:54,469 --> 00:08:57,710
Speaker 2: is large. And therefore we are in a bit of

162
00:08:57,719 --> 00:09:01,219
Speaker 2: a race to see how quickly we can build out.

163
00:09:01,570 --> 00:09:05,890
Speaker 2: Um the smarter and smarter supercomputer for lack of a

164
00:09:05,900 --> 00:09:09,609
Speaker 2: better word. And that scale is daunting. When you look

165
00:09:09,619 --> 00:09:12,329
Speaker 2: at the financial statements, you think about how are we

166
00:09:12,340 --> 00:09:16,030
Speaker 2: going to absorb the Capex investment having been a software

167
00:09:16,039 --> 00:09:17,460
Speaker 2: company for so long

168
00:09:17,750 --> 00:09:20,559
Speaker 2: that is really built on opex. We are now an

169
00:09:20,690 --> 00:09:24,108
Speaker 2: opex and Capex company and yet I look at this

170
00:09:24,119 --> 00:09:27,409
Speaker 2: as a personal problem is really that we have the

171
00:09:27,419 --> 00:09:30,130
Speaker 2: benefit of being one of those few companies that will

172
00:09:30,140 --> 00:09:34,929
Speaker 2: spend that Capex in order to deliver services that we

173
00:09:34,940 --> 00:09:39,049
Speaker 2: can't even foresee in the future that will deliver value.

174
00:09:39,359 --> 00:09:41,489
Speaker 2: Um One way of putting it. And this is something

175
00:09:41,500 --> 00:09:43,709
Speaker 2: that Bill Gates said, I think originally, which is the

176
00:09:43,719 --> 00:09:46,469
Speaker 2: value of a platform needs to be 10 times greater

177
00:09:46,479 --> 00:09:49,739
Speaker 2: for our partners than the value that we create, capture

178
00:09:49,750 --> 00:09:51,940
Speaker 2: from it. And in so many ways, I think that the,

179
00:09:51,950 --> 00:09:54,619
Speaker 2: the value that or the investment that we're putting into

180
00:09:54,630 --> 00:09:59,859
Speaker 2: building this platform should clearly unlock significantly more than 10

181
00:09:59,869 --> 00:10:03,510
Speaker 2: times the value when you think about the broader economy.

182
00:10:03,520 --> 00:10:06,780
Speaker 2: Whereas previous computing waves may have been limited

183
00:10:07,000 --> 00:10:11,239
Speaker 2: to more of an it sector, the the thinking about

184
00:10:11,250 --> 00:10:14,539
Speaker 2: and nobody really knows about the addressable market for um

185
00:10:14,549 --> 00:10:17,650
Speaker 2: the ways in which A I will unlock value in

186
00:10:17,659 --> 00:10:21,419
Speaker 2: every sector are significantly greater. So as much as this

187
00:10:21,429 --> 00:10:26,159
Speaker 2: Capex looks daunting it, it truly is what's required for

188
00:10:26,169 --> 00:10:28,159
Speaker 2: us to be able to deliver the value in the

189
00:10:28,169 --> 00:10:30,260
Speaker 2: back half of that decade. I spoke of,

190
00:10:30,770 --> 00:10:31,179
Speaker 1: right.

191
00:10:31,520 --> 00:10:33,710
Speaker 1: Um Yeah, uh allow me to go on a bit

192
00:10:33,719 --> 00:10:36,479
Speaker 1: of a tangent. It is related to this issue of

193
00:10:36,489 --> 00:10:40,260
Speaker 1: the very high cost of rolling out gen A I models,

194
00:10:40,489 --> 00:10:42,260
Speaker 1: which is the cost of innovation.

195
00:10:42,690 --> 00:10:45,710
Speaker 1: Uh You know, in, in economics, we, we have this

196
00:10:45,719 --> 00:10:49,119
Speaker 1: long standing sort of, you know, almost comical observation that

197
00:10:49,130 --> 00:10:52,750
Speaker 1: technology is everywhere except in productivity that from the 19

198
00:10:52,760 --> 00:10:55,409
Speaker 1: seventies onward, we've had a bit of a downshift in

199
00:10:55,419 --> 00:10:58,900
Speaker 1: productivity few times here and there, perhaps around the dawn

200
00:10:58,909 --> 00:11:01,890
Speaker 1: of the internet age age, we saw some pickup in productivity.

201
00:11:01,900 --> 00:11:05,440
Speaker 1: But by and large in the industrial world productivity has

202
00:11:05,450 --> 00:11:09,859
Speaker 1: been disappointing. And one argument among many arguments is that

203
00:11:09,869 --> 00:11:11,270
Speaker 1: it's just that, you know, we have sort of

204
00:11:11,690 --> 00:11:15,700
Speaker 1: exhausted the low harvesting fruits and now any tangible innovation

205
00:11:15,719 --> 00:11:18,960
Speaker 1: is very, very costly. Uh Does it have to be

206
00:11:18,969 --> 00:11:21,260
Speaker 1: like that? I mean, do you have to always spend

207
00:11:21,390 --> 00:11:23,468
Speaker 1: hundreds of billions of dollars to bring prosperity to the

208
00:11:23,479 --> 00:11:26,098
Speaker 1: world going forward because we're gonna scrape the bottom of

209
00:11:26,109 --> 00:11:26,989
Speaker 1: the barrel already.

210
00:11:28,030 --> 00:11:31,210
Speaker 2: I don't think so. I think that this particular roll

211
00:11:31,219 --> 00:11:36,368
Speaker 2: out is at a infrastructural layer which then enables people

212
00:11:36,380 --> 00:11:39,689
Speaker 2: who are building on top of the platform to use

213
00:11:39,700 --> 00:11:44,929
Speaker 2: less money to deliver productivity gain. Um I I just

214
00:11:44,940 --> 00:11:47,869
Speaker 2: saw this morning an example of how we can take

215
00:11:47,880 --> 00:11:51,369
Speaker 2: people who have lower no vision and give them through

216
00:11:51,380 --> 00:11:53,409
Speaker 2: the tools that we have with GP T four and

217
00:11:53,419 --> 00:11:54,090
Speaker 2: beyond

218
00:11:54,330 --> 00:11:56,869
Speaker 2: the ability to interact and see the world, unlocking a

219
00:11:56,880 --> 00:12:00,500
Speaker 2: lot of that productivity gain that doesn't require very much

220
00:12:00,510 --> 00:12:03,348
Speaker 2: in terms of additional cost on top of it. So

221
00:12:03,359 --> 00:12:07,840
Speaker 2: once we've built out the railroad or the interstate system

222
00:12:07,849 --> 00:12:11,669
Speaker 2: or whatever it might be that creates the fluidity in

223
00:12:11,679 --> 00:12:15,010
Speaker 2: the economy, then the productivity unlock I think happens in

224
00:12:15,020 --> 00:12:19,150
Speaker 2: terms of the innovation in every sector. So in financial services,

225
00:12:19,159 --> 00:12:21,590
Speaker 2: it might be, you know, better Ekyc

226
00:12:21,909 --> 00:12:24,979
Speaker 2: or uh micro moment lending. It might be a deeper

227
00:12:24,989 --> 00:12:27,729
Speaker 2: understanding of thin file customers. Now I'll use my first

228
00:12:27,739 --> 00:12:30,969
Speaker 2: inclusion example, which is a company called Trusting Social. A

229
00:12:30,979 --> 00:12:34,049
Speaker 2: partner of ours is thinking in Vietnam and beyond about

230
00:12:34,059 --> 00:12:37,770
Speaker 2: how it might deliver those micro moments of credit through

231
00:12:38,090 --> 00:12:42,080
Speaker 2: a generation of A I interface time war that gives

232
00:12:42,090 --> 00:12:44,890
Speaker 2: us a deeper understanding about where we can do lending

233
00:12:44,900 --> 00:12:48,598
Speaker 2: or provide protection products in the moment that's significantly better

234
00:12:48,609 --> 00:12:48,940
Speaker 2: than the way

235
00:12:49,030 --> 00:12:51,609
Speaker 2: we've done it in the past. And I think that

236
00:12:51,619 --> 00:12:56,820
Speaker 2: that ability to create value through productivity or thinking differently

237
00:12:56,830 --> 00:13:01,890
Speaker 2: about our problem is where the initial investment um will

238
00:13:01,900 --> 00:13:04,229
Speaker 2: seem like, OK, now, I understand why we built that.

239
00:13:04,239 --> 00:13:05,569
Speaker 2: And now the things that we need to do to

240
00:13:05,580 --> 00:13:07,539
Speaker 2: unlock in the future are going to be less. But

241
00:13:07,549 --> 00:13:10,728
Speaker 2: I I your point is extremely well taken, which is,

242
00:13:10,909 --> 00:13:14,650
Speaker 2: it still seems like we haven't gotten to the point

243
00:13:14,659 --> 00:13:16,280
Speaker 2: where productivity is

244
00:13:16,770 --> 00:13:20,390
Speaker 2: accelerating. In fact, you might argue that it has been

245
00:13:20,400 --> 00:13:22,789
Speaker 2: decelerating and that does go back to my earlier point

246
00:13:22,799 --> 00:13:25,939
Speaker 2: about 2000 to 2015. I shrug when I think about

247
00:13:25,950 --> 00:13:31,260
Speaker 2: the true technological events in science and engineering. But since

248
00:13:31,270 --> 00:13:36,390
Speaker 2: about 2015 to this moment, these are really the golden years.

249
00:13:36,609 --> 00:13:40,070
Speaker 2: And I'm very keen to talk about the idea around

250
00:13:40,200 --> 00:13:43,549
Speaker 2: gen A I in science and in particular around the

251
00:13:43,559 --> 00:13:45,669
Speaker 2: idea of discovery and design.

252
00:13:46,000 --> 00:13:49,559
Speaker 2: We don't know how the design of new things will

253
00:13:49,570 --> 00:13:53,739
Speaker 2: improve as a result of the discoveries that Gen A

254
00:13:53,750 --> 00:13:57,059
Speaker 2: I now enables us to make. And a simple example

255
00:13:57,070 --> 00:13:59,260
Speaker 2: was around batteries. And if you don't mind, I'll, I'll

256
00:13:59,270 --> 00:14:01,130
Speaker 2: share this story even though it was a bit of

257
00:14:01,140 --> 00:14:05,059
Speaker 2: AAA Microsoft story, which is that we were working um

258
00:14:05,070 --> 00:14:06,020
Speaker 2: on

259
00:14:06,349 --> 00:14:10,959
Speaker 2: how we could potentially figure out new battery discoveries and

260
00:14:10,969 --> 00:14:15,210
Speaker 2: that would have taken years. But fundamentally with the PNNL,

261
00:14:15,219 --> 00:14:18,309
Speaker 2: which is in, um, Washington State, we are testing a

262
00:14:18,320 --> 00:14:21,549
Speaker 2: new battery material, right? And a fundamentally what we were

263
00:14:21,559 --> 00:14:24,369
Speaker 2: thinking about doing was thinking about, um,

264
00:14:24,950 --> 00:14:29,340
Speaker 2: 5 32 million different possibilities. The A I narrowed it

265
00:14:29,349 --> 00:14:35,270
Speaker 2: down to about about um, 500,000 stable materials and then

266
00:14:35,280 --> 00:14:39,940
Speaker 2: narrowed it down even further to 18 promising candidates. So fundamentally,

267
00:14:39,950 --> 00:14:42,400
Speaker 2: it did all that trial for us

268
00:14:42,760 --> 00:14:46,090
Speaker 2: to go from 32 million to 18. And then those

269
00:14:46,099 --> 00:14:48,500
Speaker 2: 18 could be built in the real world in the

270
00:14:48,510 --> 00:14:51,299
Speaker 2: atomic world and then tested. And then one of them

271
00:14:51,309 --> 00:14:55,900
Speaker 2: turned into a breakthrough battery technology that will then design

272
00:14:56,070 --> 00:14:58,450
Speaker 2: something that will help unlock out

273
00:14:58,989 --> 00:15:01,710
Speaker 2: and that could have been yours. But it's now happened

274
00:15:01,719 --> 00:15:05,010
Speaker 2: in days as a result of the A I basically saying, no,

275
00:15:05,020 --> 00:15:07,760
Speaker 2: we're just gonna, you know, push our way through this

276
00:15:07,770 --> 00:15:12,719
Speaker 2: particular battery design and our competitors at Google uh with

277
00:15:12,729 --> 00:15:15,289
Speaker 2: the deep mind effort have always been one of the

278
00:15:15,299 --> 00:15:17,479
Speaker 2: um the the shining lights in terms of the use

279
00:15:17,489 --> 00:15:21,549
Speaker 2: case of protein folding salt significantly faster than we all thought.

280
00:15:21,619 --> 00:15:23,950
Speaker 2: And we and a number of the companies are working

281
00:15:23,960 --> 00:15:26,429
Speaker 2: at the protein level and then even at a a

282
00:15:26,440 --> 00:15:28,570
Speaker 2: nano level to figure out how we might design

283
00:15:28,960 --> 00:15:32,539
Speaker 2: um new drugs and new therapies that will improve health outcomes.

284
00:15:32,549 --> 00:15:35,380
Speaker 2: So I go back to this idea that the productivity

285
00:15:35,390 --> 00:15:38,750
Speaker 2: gain that we might see in all the other fields

286
00:15:38,760 --> 00:15:41,289
Speaker 2: when it comes from energy or education, from banking to

287
00:15:41,299 --> 00:15:45,090
Speaker 2: health care are unlocked by this $200 billion of Capex

288
00:15:45,099 --> 00:15:45,859
Speaker 2: that we're going to see

289
00:15:46,000 --> 00:15:49,840
Speaker 2: hit um the financial statements of companies like ours for

290
00:15:49,849 --> 00:15:52,020
Speaker 2: the next little while. But we see this happen in

291
00:15:52,030 --> 00:15:55,469
Speaker 2: history before and it has worked before and I'm not

292
00:15:55,479 --> 00:15:58,539
Speaker 2: just being an optimist. I truly do believe that as

293
00:15:58,549 --> 00:16:00,700
Speaker 2: a tool. This is something that has been greater than

294
00:16:00,710 --> 00:16:02,219
Speaker 2: many of the other things that we've seen in a

295
00:16:02,229 --> 00:16:02,940
Speaker 2: long time.

296
00:16:03,450 --> 00:16:07,210
Speaker 1: So it came to Eisenhower's Interstate Highway program in the

297
00:16:07,219 --> 00:16:10,570
Speaker 1: fifties or laying submarine cables to connect the world through

298
00:16:10,580 --> 00:16:13,039
Speaker 1: internet in the nineties and two thousands.

299
00:16:13,559 --> 00:16:16,559
Speaker 2: Yeah, I think that it, those are, those are good metaphors.

300
00:16:16,570 --> 00:16:20,880
Speaker 2: Um And uh I, I think that the railways is

301
00:16:20,890 --> 00:16:24,789
Speaker 2: also a good metaphor and, and any other infrastructure, the

302
00:16:24,799 --> 00:16:28,849
Speaker 2: internet itself, you know, the dest development and, and to

303
00:16:28,859 --> 00:16:30,599
Speaker 2: go back to the printing press because it is a

304
00:16:30,609 --> 00:16:33,659
Speaker 2: tool right in the hands of humans to do things

305
00:16:33,669 --> 00:16:39,179
Speaker 2: that we could not predict and education. And um and,

306
00:16:39,190 --> 00:16:41,500
Speaker 2: and maybe even art

307
00:16:41,799 --> 00:16:44,940
Speaker 2: is, is super interesting and how it might be unlocked,

308
00:16:44,950 --> 00:16:47,059
Speaker 2: uh which we didn't think of two years ago that

309
00:16:47,070 --> 00:16:47,729
Speaker 2: the creative

310
00:16:47,739 --> 00:16:48,309
Speaker 2: side of it.

311
00:16:48,809 --> 00:16:50,750
Speaker 1: Right. Uh Yeah, I mean, I, as you know, I

312
00:16:50,760 --> 00:16:54,469
Speaker 1: have an 11 year old, I'm excited about an infinitely patient.

313
00:16:54,479 --> 00:16:57,820
Speaker 1: A I became a hard math skills which, you know,

314
00:16:57,830 --> 00:17:01,140
Speaker 1: be unmatched by any teachers. Uh even the most patient teacher,

315
00:17:01,150 --> 00:17:03,179
Speaker 1: that's what I'm looking forward to on the education side

316
00:17:03,190 --> 00:17:07,329
Speaker 1: of uh Ja I application. Uh Yeah, tell me about

317
00:17:07,339 --> 00:17:09,208
Speaker 1: the five phases of gen A I.

318
00:17:09,958 --> 00:17:12,068
Speaker 2: Great thanks. And I don't think this is necessarily the

319
00:17:12,078 --> 00:17:16,158
Speaker 2: five phases perhaps, but it is again a way of

320
00:17:16,167 --> 00:17:19,338
Speaker 2: thinking about this decade that we might be in and

321
00:17:19,678 --> 00:17:23,198
Speaker 2: it has five steps to it. And without sharing this

322
00:17:23,208 --> 00:17:26,389
Speaker 2: on the screen, I think people can visualize something that,

323
00:17:26,398 --> 00:17:30,409
Speaker 2: which looks like an exponential curve where uh the X

324
00:17:30,417 --> 00:17:31,409
Speaker 2: axis is time

325
00:17:31,760 --> 00:17:35,109
Speaker 2: or some sort of level of um of automation. And

326
00:17:35,119 --> 00:17:37,050
Speaker 2: then the Y axis is value like just how much

327
00:17:37,060 --> 00:17:39,780
Speaker 2: value do we deliver from it. And we are still

328
00:17:39,790 --> 00:17:42,099
Speaker 2: in like I can call it a 532 model, but

329
00:17:42,109 --> 00:17:44,599
Speaker 2: we're still in that first five years, we're in the

330
00:17:44,609 --> 00:17:46,849
Speaker 2: current phase of a

331
00:17:46,935 --> 00:17:50,435
Speaker 2: uh the first two steps, there are conversational A I

332
00:17:50,625 --> 00:17:54,214
Speaker 2: and maybe business knowledge, then we will move into the

333
00:17:54,224 --> 00:17:58,675
Speaker 2: 2nd 3rd, if you will, where skills for work and

334
00:17:58,685 --> 00:18:01,313
Speaker 2: then sort of reasoning and problem solving will start to

335
00:18:01,324 --> 00:18:01,954
Speaker 2: kick in.

336
00:18:02,369 --> 00:18:05,219
Speaker 2: And then finally, we'll get to the, the last phase

337
00:18:05,229 --> 00:18:08,209
Speaker 2: where we'll get self learning. Now, I'll step through these

338
00:18:08,219 --> 00:18:11,169
Speaker 2: pretty quickly because, you know, I just, I recognize as

339
00:18:11,180 --> 00:18:14,589
Speaker 2: a podcast and not a presentation, but look, the interfaces

340
00:18:14,599 --> 00:18:18,339
Speaker 2: of the past have usually been mechanical or punch cards, keyboards,

341
00:18:18,349 --> 00:18:21,750
Speaker 2: gesture interfaces like the mouse trackpad, touch screen. We are

342
00:18:21,760 --> 00:18:25,948
Speaker 2: already starting to see ja I use conversational interfaces that

343
00:18:25,959 --> 00:18:27,989
Speaker 2: are more natural and more human like.

344
00:18:28,114 --> 00:18:30,724
Speaker 2: And in many ways, we in Asia have known this

345
00:18:30,734 --> 00:18:34,944
Speaker 2: for quite some time. It isn't just about typing into

346
00:18:34,964 --> 00:18:38,405
Speaker 2: a window and then generating a response. We want our

347
00:18:38,415 --> 00:18:44,135
Speaker 2: conversational A I to be Multilingual, multimodal, spoken in one language,

348
00:18:44,145 --> 00:18:47,415
Speaker 2: you know, responded to it in another um over an

349
00:18:47,425 --> 00:18:52,844
Speaker 2: Android phone, over a weak connection that's underpowered um on

350
00:18:52,854 --> 00:18:53,764
Speaker 2: the back of a motorbike.

351
00:18:54,109 --> 00:18:57,420
Speaker 2: So the ways in which we in Asia demand A

352
00:18:57,430 --> 00:18:59,889
Speaker 2: I to be used so that it is applicable and

353
00:18:59,900 --> 00:19:05,770
Speaker 2: useful require a conversational approach and not full literacy or

354
00:19:05,780 --> 00:19:08,969
Speaker 2: numeracy in English. And so we're getting there and we

355
00:19:08,979 --> 00:19:10,339
Speaker 2: saw a couple of things that have happened the last

356
00:19:10,349 --> 00:19:12,849
Speaker 2: couple of years, a couple of days even or weeks

357
00:19:12,859 --> 00:19:14,729
Speaker 2: around how we've tried to think about this to be

358
00:19:14,739 --> 00:19:15,849
Speaker 2: more conversational.

359
00:19:16,849 --> 00:19:19,639
Speaker 2: So how will it help? I mean, just basically makes

360
00:19:19,650 --> 00:19:23,030
Speaker 2: it more accessible. Uh Someone once said English is the

361
00:19:23,040 --> 00:19:27,290
Speaker 2: hottest new programming language. Where does this go is really

362
00:19:27,300 --> 00:19:30,939
Speaker 2: in terms of the the idea around workforce, right? And

363
00:19:30,949 --> 00:19:32,930
Speaker 2: the idea is that we are using this as a

364
00:19:32,939 --> 00:19:37,469
Speaker 2: tool to improve our productivity. But in many ways, our

365
00:19:37,479 --> 00:19:38,140
Speaker 2: time

366
00:19:38,469 --> 00:19:42,020
Speaker 2: hours of of the day are spent on doing tasks

367
00:19:42,030 --> 00:19:45,579
Speaker 2: that can be routinized or can be done to summarize,

368
00:19:45,589 --> 00:19:50,839
Speaker 2: to analyze, to a log uh to shorten creation time.

369
00:19:50,849 --> 00:19:53,359
Speaker 2: And this idea that we're doing some of the same things,

370
00:19:53,369 --> 00:19:55,349
Speaker 2: but in shorter periods of time

371
00:19:55,435 --> 00:19:58,594
Speaker 2: or with the assistance of some other type of tool,

372
00:19:58,604 --> 00:20:01,775
Speaker 2: perhaps a little bit like a calculator. If you think

373
00:20:01,785 --> 00:20:03,964
Speaker 2: of it that way, or another type of tool in

374
00:20:03,974 --> 00:20:08,055
Speaker 2: whatever field that you're in, basically transforms and accelerates and

375
00:20:08,064 --> 00:20:11,353
Speaker 2: assists the way we work to succeed in the modern

376
00:20:11,364 --> 00:20:12,294
Speaker 2: work face, right?

377
00:20:12,550 --> 00:20:15,239
Speaker 2: But then as we go into the phases three and four,

378
00:20:15,250 --> 00:20:18,000
Speaker 2: what we really get into is this idea that you

379
00:20:18,010 --> 00:20:22,208
Speaker 2: get domain specific skills for work. And that is when

380
00:20:22,219 --> 00:20:26,129
Speaker 2: an agent maybe takes over some of the things that

381
00:20:26,140 --> 00:20:29,149
Speaker 2: you might have had a group of humans or tools

382
00:20:29,430 --> 00:20:33,969
Speaker 2: do and that agent can act on behalf of a company,

383
00:20:33,979 --> 00:20:37,810
Speaker 2: a function or an individual to perform a task. It

384
00:20:37,819 --> 00:20:41,369
Speaker 2: could be something like credit scoring or, or, or underwriting.

385
00:20:41,380 --> 00:20:44,589
Speaker 2: It could be something as simple as as uh organizing

386
00:20:44,599 --> 00:20:47,619
Speaker 2: a set of information and those skills for work will

387
00:20:47,630 --> 00:20:50,229
Speaker 2: then start to be more advanced in terms of how

388
00:20:50,239 --> 00:20:54,680
Speaker 2: they might reorganize the or chart. For example,

389
00:20:54,930 --> 00:20:56,948
Speaker 2: it gets exciting for me in that fourth phase where

390
00:20:56,959 --> 00:20:59,160
Speaker 2: we start to do planning. And I don't mean we,

391
00:20:59,170 --> 00:21:02,209
Speaker 2: but I just generally think that the tools start to

392
00:21:02,219 --> 00:21:04,849
Speaker 2: do planning about what it is that they're gonna need

393
00:21:04,859 --> 00:21:08,670
Speaker 2: to learn in order to solve the problem up till now,

394
00:21:08,680 --> 00:21:11,349
Speaker 2: we've given them a problem. If they have that skill, they, they,

395
00:21:11,359 --> 00:21:13,170
Speaker 2: they basically figure out how to do it faster than

396
00:21:13,180 --> 00:21:16,079
Speaker 2: we did, you know, 99 faster than 99% of humans.

397
00:21:16,219 --> 00:21:17,670
Speaker 2: But when you give them the ability

398
00:21:17,770 --> 00:21:20,069
Speaker 2: to plan out and figure out how they learn and

399
00:21:20,079 --> 00:21:23,599
Speaker 2: reason in order to solve an open ended problem, then

400
00:21:23,609 --> 00:21:25,750
Speaker 2: that's when you start, start to get to more of

401
00:21:25,760 --> 00:21:28,560
Speaker 2: an advanced phase. And this is again, phase four of

402
00:21:28,569 --> 00:21:32,800
Speaker 2: reasoning and problem solving. Finally, the last phase is self

403
00:21:32,810 --> 00:21:36,010
Speaker 2: learning is when people, when, when the agents and when

404
00:21:36,020 --> 00:21:39,790
Speaker 2: the artificial intelligence learns new skills that we didn't tell

405
00:21:39,800 --> 00:21:40,430
Speaker 2: it to learn,

406
00:21:41,420 --> 00:21:43,698
Speaker 2: not just because we gave it a task, but because

407
00:21:43,709 --> 00:21:47,160
Speaker 2: it realizes that it can and it develops a more

408
00:21:47,170 --> 00:21:50,920
Speaker 2: generalized intelligence. So the G moves from the, from the

409
00:21:50,930 --> 00:21:53,780
Speaker 2: beginning of the word to the center of the world

410
00:21:53,790 --> 00:21:57,479
Speaker 2: and we enter into the artificial general intelligence space or

411
00:21:57,704 --> 00:22:01,415
Speaker 2: G I and that will unlock potentially a lot of

412
00:22:01,425 --> 00:22:05,083
Speaker 2: different ways in which the marginal cost of many things

413
00:22:05,094 --> 00:22:08,324
Speaker 2: drops down to close to zero. And I'm a big

414
00:22:08,334 --> 00:22:11,405
Speaker 2: believer that humanity will still persist even in the presence

415
00:22:11,415 --> 00:22:12,794
Speaker 2: of A G I. There's one thing I'm going to

416
00:22:12,805 --> 00:22:15,544
Speaker 2: end with is that when we even get to that

417
00:22:15,555 --> 00:22:18,794
Speaker 2: point where we have processes and tools and agents that

418
00:22:18,805 --> 00:22:20,755
Speaker 2: do some of the more menial things,

419
00:22:21,199 --> 00:22:24,438
Speaker 2: um We'll be outsourcing these things to A G I

420
00:22:24,449 --> 00:22:28,540
Speaker 2: but still valuing time more and unique human abilities. We

421
00:22:28,550 --> 00:22:32,540
Speaker 2: may have a re appreciation for craft, for artisan recognition

422
00:22:32,550 --> 00:22:36,250
Speaker 2: and reward for creation, empathy and mental, mental and physical

423
00:22:36,260 --> 00:22:39,169
Speaker 2: health will spike. People will still need to seek help

424
00:22:39,180 --> 00:22:44,339
Speaker 2: and seek support. We love competition, we love love and

425
00:22:44,349 --> 00:22:46,839
Speaker 2: the enjoyment of drama. I think that we're going to

426
00:22:46,849 --> 00:22:48,010
Speaker 2: have an adaptation of a

427
00:22:48,055 --> 00:22:51,224
Speaker 2: solution of what it means to frankly, to be human

428
00:22:51,234 --> 00:22:53,675
Speaker 2: and maybe our, our standards for what we want out

429
00:22:53,685 --> 00:22:57,594
Speaker 2: of humanity will rise in 10 years. Another way of

430
00:22:57,604 --> 00:22:59,474
Speaker 2: putting it is that the things that we thought were

431
00:22:59,484 --> 00:23:03,435
Speaker 2: just sort of very difficult to do 2030 years, maybe

432
00:23:03,444 --> 00:23:06,165
Speaker 2: not in our lifetime or our careers. Now, all of

433
00:23:06,175 --> 00:23:07,675
Speaker 2: a sudden they could happen

434
00:23:08,349 --> 00:23:11,639
Speaker 2: 57, 10 years earlier and we might be living in

435
00:23:11,650 --> 00:23:14,708
Speaker 2: a time where we solved nuclear fusion. One of my

436
00:23:14,719 --> 00:23:18,250
Speaker 2: favorites either through a new Telcom Ma reactor design or

437
00:23:18,260 --> 00:23:21,650
Speaker 2: through some other types of means we accelerate the process

438
00:23:21,660 --> 00:23:23,770
Speaker 2: by which we come up with the discoveries we need

439
00:23:23,780 --> 00:23:26,429
Speaker 2: in climate tech to solve some of the issues around

440
00:23:26,439 --> 00:23:28,079
Speaker 2: energy sooner and that,

441
00:23:28,324 --> 00:23:31,063
Speaker 2: that optimism that we unlock some of these things that

442
00:23:31,074 --> 00:23:34,604
Speaker 2: we don't even know about yet. Through A I and

443
00:23:34,614 --> 00:23:37,354
Speaker 2: through the combination of humans using A I in interesting

444
00:23:37,364 --> 00:23:39,925
Speaker 2: ways in each of these verticals, gets us to that

445
00:23:39,935 --> 00:23:45,265
Speaker 2: point where we're solving big, big problems like education, like health,

446
00:23:45,275 --> 00:23:48,665
Speaker 2: health outcomes like inclusivity and of course, like energy uh

447
00:23:48,675 --> 00:23:51,464
Speaker 2: renewable energy sooner rather than later.

448
00:23:52,599 --> 00:23:55,020
Speaker 1: Sure. And you know, when you were talking about the

449
00:23:55,030 --> 00:23:57,930
Speaker 1: final phases of A I, I was thinking about Iron

450
00:23:57,939 --> 00:24:02,369
Speaker 1: Man and how Tony Stark interacts with Jarvis uh where

451
00:24:02,410 --> 00:24:06,900
Speaker 1: Jarvis is not necessarily all knowing but is an incredible

452
00:24:06,910 --> 00:24:08,170
Speaker 1: complement to Tony's

453
00:24:08,430 --> 00:24:12,949
Speaker 1: brilliance and, and helping him solve problems in, in speed.

454
00:24:12,959 --> 00:24:16,869
Speaker 1: That is not humanly possible. So I'm not really sure

455
00:24:16,880 --> 00:24:19,300
Speaker 1: Jarvis needs to have a G I or a very

456
00:24:19,310 --> 00:24:21,589
Speaker 1: advanced form of J I is good enough. But uh

457
00:24:21,599 --> 00:24:23,089
Speaker 1: but that's what I was thinking about when you were

458
00:24:23,099 --> 00:24:23,819
Speaker 1: talking about that.

459
00:24:24,260 --> 00:24:26,010
Speaker 2: You read one of my previous slides, I have a

460
00:24:26,020 --> 00:24:29,209
Speaker 2: slide that says, and it's got, it's got Tony Stark

461
00:24:29,219 --> 00:24:31,260
Speaker 2: saying it will be more like Jarvis and will be

462
00:24:31,270 --> 00:24:32,239
Speaker 2: like terminator.

463
00:24:32,969 --> 00:24:35,369
Speaker 2: And so you, you, you read my mind on that

464
00:24:35,380 --> 00:24:37,639
Speaker 2: one and I'll throw out to the team like I

465
00:24:37,650 --> 00:24:42,239
Speaker 2: love to end this particular phase of this discussion around genius.

466
00:24:42,599 --> 00:24:46,030
Speaker 2: All right. And in, in, I'll ask the audience to,

467
00:24:46,040 --> 00:24:48,260
Speaker 2: to think about what is the next best word that

468
00:24:48,270 --> 00:24:51,369
Speaker 2: comes to their mind. When I start with this particular line,

469
00:24:51,430 --> 00:24:54,880
Speaker 2: the first two words are misery loves

470
00:24:56,199 --> 00:24:57,170
Speaker 1: his company,

471
00:24:57,229 --> 00:24:59,159
Speaker 2: his company. So I've done this in a variety of

472
00:24:59,170 --> 00:25:02,400
Speaker 2: contexts and some context of younger people don't know what

473
00:25:02,410 --> 00:25:04,849
Speaker 2: the word might be. But the suggestion that it is

474
00:25:04,859 --> 00:25:08,250
Speaker 2: company is a moment of genius

475
00:25:08,599 --> 00:25:12,319
Speaker 2: from when it was first written in 1600 by William Shakespeare.

476
00:25:12,380 --> 00:25:16,909
Speaker 2: And since then any tool, any reference, even any A

477
00:25:16,920 --> 00:25:20,969
Speaker 2: I is just holding a mirror up onto what it

478
00:25:20,979 --> 00:25:25,369
Speaker 2: is that we have done and saying that the next best,

479
00:25:25,380 --> 00:25:29,380
Speaker 2: most likely word for Misery Loves is company. But until

480
00:25:29,760 --> 00:25:33,260
Speaker 2: the Great Bard wrote that word, it would not be

481
00:25:33,270 --> 00:25:38,099
Speaker 2: suggested amongst the 10,000 English words as the next best answer.

482
00:25:38,109 --> 00:25:41,079
Speaker 2: So we will still have moments of genius that will

483
00:25:41,089 --> 00:25:45,069
Speaker 2: then embed themselves into the corpus from which the A

484
00:25:45,079 --> 00:25:48,140
Speaker 2: I can then say maybe the next word is company.

485
00:25:49,280 --> 00:25:53,060
Speaker 1: Very good. Uh Zia, you talked about one of your

486
00:25:53,069 --> 00:25:57,479
Speaker 1: dreams being, you know, uh clean nuclear fusion and so on. Well,

487
00:25:57,489 --> 00:26:00,770
Speaker 1: we're not there yet and therefore, right now, we're still

488
00:26:00,780 --> 00:26:04,780
Speaker 1: burning a lot of fossil fuel and some renewables to

489
00:26:05,020 --> 00:26:09,020
Speaker 1: power our current and aspiring world. So let's talk about

490
00:26:09,030 --> 00:26:12,060
Speaker 1: the compute needs and the carbon footprint of running gen

491
00:26:12,069 --> 00:26:12,780
Speaker 1: A I models.

492
00:26:14,489 --> 00:26:18,188
Speaker 2: There's no doubt that they are high and that the

493
00:26:18,219 --> 00:26:23,569
Speaker 2: potential for them to grow at less fast a pace

494
00:26:23,579 --> 00:26:25,109
Speaker 2: is clearly there.

495
00:26:25,680 --> 00:26:28,389
Speaker 2: So we care very much at Microsoft about our scope

496
00:26:28,400 --> 00:26:30,530
Speaker 2: three emissions. Of course, scope one and two is super

497
00:26:30,540 --> 00:26:33,500
Speaker 2: important and those have dropped. But scope three continues to

498
00:26:33,510 --> 00:26:36,140
Speaker 2: rise and it is as a result of the fact

499
00:26:36,150 --> 00:26:39,379
Speaker 2: that many of the places in which we we we

500
00:26:39,390 --> 00:26:42,419
Speaker 2: get components for the supercomputers and the things that we

501
00:26:42,430 --> 00:26:46,560
Speaker 2: need to build are still using grids that are based

502
00:26:46,569 --> 00:26:51,219
Speaker 2: on non renewable energy. And in particular, more than 50%

503
00:26:51,229 --> 00:26:54,030
Speaker 2: of most companies scope three emissions actually come from right

504
00:26:54,040 --> 00:26:55,219
Speaker 2: here right here in Asia.

505
00:26:55,569 --> 00:26:58,609
Speaker 2: And as you look at the different grids in China

506
00:26:58,619 --> 00:27:01,510
Speaker 2: is doing super well. Um We are starting to see

507
00:27:01,520 --> 00:27:05,260
Speaker 2: an inflection point when it comes to renewable energy. And

508
00:27:05,270 --> 00:27:09,430
Speaker 2: um recently, we've seen in the West and in countries

509
00:27:09,439 --> 00:27:12,750
Speaker 2: like China, um that tipping point where the cost of

510
00:27:12,760 --> 00:27:15,270
Speaker 2: renewable energy gets to the point where it starts to

511
00:27:15,280 --> 00:27:16,589
Speaker 2: be more affordable,

512
00:27:16,810 --> 00:27:19,520
Speaker 2: but we're by no means anywhere close to that point

513
00:27:19,530 --> 00:27:22,488
Speaker 2: yet for the rest of Asia. And so what we

514
00:27:22,500 --> 00:27:26,030
Speaker 2: are hoping to do is to ask our partners in

515
00:27:26,040 --> 00:27:29,560
Speaker 2: markets where we need to build a data center to

516
00:27:29,569 --> 00:27:34,639
Speaker 2: think about the ways in which renewable energy can unlock

517
00:27:34,660 --> 00:27:37,680
Speaker 2: the decision to put a data center nearby.

518
00:27:38,239 --> 00:27:41,109
Speaker 2: Put it differently if you were making a decision about

519
00:27:41,119 --> 00:27:42,900
Speaker 2: where to put a data center that's gonna suck up

520
00:27:42,910 --> 00:27:45,839
Speaker 2: some of the earth's resources. And there was no sovereign

521
00:27:45,849 --> 00:27:49,140
Speaker 2: data issue and we had zero latency and we had

522
00:27:49,150 --> 00:27:52,209
Speaker 2: the ability to transport electrons from one part of the

523
00:27:52,219 --> 00:27:54,609
Speaker 2: world to the other but not electricity.

524
00:27:55,390 --> 00:28:00,300
Speaker 2: You would always locate your data centers next to a

525
00:28:00,310 --> 00:28:04,949
Speaker 2: place that could generate uh energy renewable and cleanly rather

526
00:28:04,959 --> 00:28:05,569
Speaker 2: than not.

527
00:28:06,469 --> 00:28:10,449
Speaker 2: And yet it is the distortions of sovereign data and

528
00:28:10,459 --> 00:28:15,129
Speaker 2: obviously geography that make us think differently. But in the

529
00:28:15,140 --> 00:28:21,688
Speaker 2: competition for where the compute and where the intelligence of

530
00:28:21,699 --> 00:28:25,780
Speaker 2: what it takes to take advantage of A I uh occurs.

531
00:28:26,060 --> 00:28:29,880
Speaker 2: The solution is find a way to create more renewable energy.

532
00:28:29,890 --> 00:28:32,329
Speaker 2: And you will find a way to be more convincing

533
00:28:32,339 --> 00:28:33,729
Speaker 2: as a destination,

534
00:28:34,150 --> 00:28:36,979
Speaker 2: not just for D CS, but for this whole A

535
00:28:36,989 --> 00:28:40,650
Speaker 2: I wa and I firmly believe that um A I

536
00:28:40,660 --> 00:28:45,550
Speaker 2: is unlocked faster with renewable energy and we are consciously

537
00:28:45,560 --> 00:28:48,170
Speaker 2: trying to hit our 2030 commitments. We and many other

538
00:28:48,180 --> 00:28:50,280
Speaker 2: companies like ours. And the only way we're going to

539
00:28:50,290 --> 00:28:54,729
Speaker 2: do that is in partnership with large public sector entities

540
00:28:55,060 --> 00:28:57,890
Speaker 2: and private sector entities that are delivering renewable energy at

541
00:28:57,900 --> 00:29:02,140
Speaker 2: scale so that we can eventually get there. I think

542
00:29:02,150 --> 00:29:03,619
Speaker 2: we may be the first company to ever sign up

543
00:29:03,630 --> 00:29:07,439
Speaker 2: a power purchase agreement with a nuclear fusion company. Um

544
00:29:07,449 --> 00:29:10,119
Speaker 2: That's starting, it's due to becoming online in 27 and,

545
00:29:10,130 --> 00:29:11,619
Speaker 2: and I I may be wrong in terms of whether

546
00:29:11,630 --> 00:29:14,199
Speaker 2: it is fusion or fission. But it's super interesting to

547
00:29:14,209 --> 00:29:17,560
Speaker 2: see that if a large buyer of an energy says

548
00:29:17,569 --> 00:29:19,729
Speaker 2: I'm only gonna buy renewable energy or a mix that's

549
00:29:19,739 --> 00:29:23,400
Speaker 2: mostly renewable, it changes behaviors, it changes economics of the market.

550
00:29:24,140 --> 00:29:28,140
Speaker 1: Yeah, for sure. Um You talked about Asia's energy needs.

551
00:29:28,150 --> 00:29:29,989
Speaker 1: Can we just talk about the US energy needs for

552
00:29:30,000 --> 00:29:32,609
Speaker 1: a second? Because my understanding is a lot of the

553
00:29:32,619 --> 00:29:35,180
Speaker 1: data centers, a lot of the compute will be in

554
00:29:35,189 --> 00:29:38,650
Speaker 1: the US. And I've been reading articles about how in

555
00:29:38,660 --> 00:29:41,050
Speaker 1: vast parts of the US energy demand is,

556
00:29:41,130 --> 00:29:43,800
Speaker 1: is outstripping supply. California is an exception. But there are

557
00:29:43,810 --> 00:29:46,969
Speaker 1: others where this issue is becoming a bit of a,

558
00:29:46,979 --> 00:29:50,290
Speaker 1: you know, conundrum for those who want to pursue as

559
00:29:50,300 --> 00:29:54,469
Speaker 1: fast expansion of gen A I related models and compute

560
00:29:54,630 --> 00:29:56,119
Speaker 1: and those who want to sort of balance it out

561
00:29:56,130 --> 00:29:58,069
Speaker 1: from a carbon footprint perspective.

562
00:29:58,739 --> 00:30:02,170
Speaker 2: Yeah, I, I'm not as familiar with the state by

563
00:30:02,180 --> 00:30:07,500
Speaker 2: state potential for renewable energy as I probably should be,

564
00:30:07,599 --> 00:30:11,079
Speaker 2: but I will go back to the idea that it

565
00:30:11,089 --> 00:30:14,319
Speaker 2: is lumpy, right? Uh The future is already here. It

566
00:30:14,329 --> 00:30:17,479
Speaker 2: just isn't evenly distributed yet. And if you think about

567
00:30:17,489 --> 00:30:21,239
Speaker 2: the simple logic of hours of sunlight

568
00:30:22,109 --> 00:30:27,000
Speaker 2: and draw that as a curve, it may closely match

569
00:30:27,010 --> 00:30:31,329
Speaker 2: the potential for data center or deployment uh with the

570
00:30:31,339 --> 00:30:35,020
Speaker 2: exception of, of, of course of wind. But fundamentally if

571
00:30:35,030 --> 00:30:37,189
Speaker 2: you look at the southern states and the ones that

572
00:30:37,199 --> 00:30:40,170
Speaker 2: have more um sunlight,

573
00:30:40,560 --> 00:30:42,339
Speaker 2: it tells you, I think it gives you a clue

574
00:30:42,349 --> 00:30:45,609
Speaker 2: as to where PV, which is actually getting more efficient

575
00:30:45,619 --> 00:30:48,579
Speaker 2: at a close to a rate if not faster than

576
00:30:48,589 --> 00:30:54,020
Speaker 2: the uh compute power required gets us there. Um And

577
00:30:54,030 --> 00:30:56,449
Speaker 2: I think that gives us, gives me promise for the

578
00:30:56,459 --> 00:31:00,630
Speaker 2: global South Time War. Um places like where you, you

579
00:31:00,640 --> 00:31:02,530
Speaker 2: and I and our parents grew up places that we

580
00:31:02,540 --> 00:31:04,939
Speaker 2: have visited where you think about wars are going to

581
00:31:04,949 --> 00:31:05,380
Speaker 2: be

582
00:31:05,619 --> 00:31:08,709
Speaker 2: potential and you know, we just announced a billion dollar

583
00:31:08,719 --> 00:31:12,119
Speaker 2: investment through G 42 in Kenya. Um And there are

584
00:31:12,130 --> 00:31:14,390
Speaker 2: places like Iran and Pakistan, there are places that are,

585
00:31:14,410 --> 00:31:17,709
Speaker 2: have many, many days of sun that could potentially turn

586
00:31:17,719 --> 00:31:18,630
Speaker 2: that into

587
00:31:19,000 --> 00:31:23,709
Speaker 2: uh future power. So answering the question simply, it's probably PV,

588
00:31:23,719 --> 00:31:27,260
Speaker 2: that is a, a clue as to where the potential

589
00:31:27,270 --> 00:31:29,869
Speaker 2: will be. But what you said about America is probably

590
00:31:29,880 --> 00:31:32,920
Speaker 2: true about every other nation in the world with few

591
00:31:32,930 --> 00:31:36,819
Speaker 2: exceptions that just simply have exceptional renewable power potential.

592
00:31:37,949 --> 00:31:42,349
Speaker 1: Um Earlier, uh you, you made that point about Jarvis

593
00:31:42,359 --> 00:31:44,939
Speaker 1: versus Terminator. I want to come back to that point

594
00:31:44,969 --> 00:31:48,130
Speaker 1: uh because there's a broader discussion about the dystopian A

595
00:31:48,140 --> 00:31:51,689
Speaker 1: I scenarios and the utopian scenarios. So what state of

596
00:31:51,699 --> 00:31:54,280
Speaker 1: future do you think about uh Zia on a day

597
00:31:54,290 --> 00:31:55,089
Speaker 1: to day basis?

598
00:31:57,369 --> 00:32:00,189
Speaker 2: I, I do believe that

599
00:32:01,319 --> 00:32:07,339
Speaker 2: we will have a guarded stepwise approach to how we

600
00:32:07,349 --> 00:32:13,000
Speaker 2: think about these bigger things around A I principles around fairness,

601
00:32:13,010 --> 00:32:17,520
Speaker 2: which I've somewhat alluded to this idea around transparency and accountability.

602
00:32:17,530 --> 00:32:21,069
Speaker 2: If you will, we, whenever we put out a product

603
00:32:21,079 --> 00:32:24,719
Speaker 2: at Microsoft, we goes through a second review around reliability

604
00:32:24,729 --> 00:32:25,660
Speaker 2: and safety.

605
00:32:26,069 --> 00:32:29,099
Speaker 2: Plus, we have more safeguards around privacy and security than

606
00:32:29,109 --> 00:32:31,810
Speaker 2: we ever have. So the principles that we have around

607
00:32:31,819 --> 00:32:32,660
Speaker 2: how we deploy

608
00:32:33,099 --> 00:32:36,680
Speaker 2: are going to act as a natural governor for how

609
00:32:36,689 --> 00:32:38,920
Speaker 2: we do it. And why we do it. And I

610
00:32:38,930 --> 00:32:41,520
Speaker 2: think that that is something that we will share with

611
00:32:41,530 --> 00:32:45,329
Speaker 2: the industry, with our partners and with other tech companies

612
00:32:45,520 --> 00:32:47,449
Speaker 2: and how we'll do there is we're going to do

613
00:32:47,459 --> 00:32:49,560
Speaker 2: skilling and we're going to do testing and tools and

614
00:32:49,569 --> 00:32:52,880
Speaker 2: training to make sure that the way it's implemented is

615
00:32:52,890 --> 00:32:56,520
Speaker 2: the right way. And of course, there's oversight around compliance

616
00:32:56,530 --> 00:32:59,040
Speaker 2: and auditing, reporting, monitoring and we're going to build tools

617
00:32:59,050 --> 00:33:01,359
Speaker 2: to help people figure out just how they're using A

618
00:33:01,369 --> 00:33:01,680
Speaker 2: I

619
00:33:02,239 --> 00:33:04,459
Speaker 2: I think though that where it's going with the root

620
00:33:04,469 --> 00:33:08,319
Speaker 2: of the question is um this idea of optimism that

621
00:33:08,329 --> 00:33:15,140
Speaker 2: the optimism in terms of how a new group of

622
00:33:15,150 --> 00:33:18,839
Speaker 2: agents that's a billion or 2 billion strong that you

623
00:33:18,849 --> 00:33:21,119
Speaker 2: don't have to pay any money. But then basically deliver

624
00:33:21,130 --> 00:33:25,630
Speaker 2: work that nobody wants to do that just simply creates

625
00:33:25,640 --> 00:33:30,380
Speaker 2: value in places that then allows us to shift

626
00:33:30,719 --> 00:33:34,219
Speaker 2: to other activities. It's really the, the, the most interesting

627
00:33:34,229 --> 00:33:36,359
Speaker 2: part for me and I will go back to health

628
00:33:36,369 --> 00:33:38,969
Speaker 2: outcomes because it's something I've thought about over the years

629
00:33:39,060 --> 00:33:41,979
Speaker 2: and my personal family issues around health and how we

630
00:33:41,989 --> 00:33:44,939
Speaker 2: might improve the way in which we look at orphan

631
00:33:44,949 --> 00:33:49,280
Speaker 2: diseases or even simple things like um early detection of cancer.

632
00:33:49,319 --> 00:33:51,520
Speaker 2: And we've seen so many companies that look at and

633
00:33:51,530 --> 00:33:54,520
Speaker 2: see the hardware infrastructure we have today when you apply

634
00:33:54,530 --> 00:33:58,609
Speaker 2: smarter software and A I to it will provide an assist.

635
00:33:58,670 --> 00:34:00,329
Speaker 2: It's a great word and assist

636
00:34:00,599 --> 00:34:05,739
Speaker 2: to a well meaning thoughtful technology oriented doctor to be

637
00:34:05,750 --> 00:34:07,680
Speaker 2: able to say, you know what, I don't see a

638
00:34:07,689 --> 00:34:11,739
Speaker 2: shadow in the JPEG version of that file if you will.

639
00:34:11,750 --> 00:34:13,820
Speaker 2: But if you look at the raw version of the

640
00:34:13,830 --> 00:34:15,840
Speaker 2: of the file of that image that was created, it

641
00:34:15,850 --> 00:34:18,850
Speaker 2: went back to like the DS LR metaphor in that

642
00:34:18,860 --> 00:34:21,620
Speaker 2: raw image 99 times out of 100. That thing that

643
00:34:21,629 --> 00:34:24,199
Speaker 2: you see is actually a shadow and A I thinks

644
00:34:24,209 --> 00:34:27,919
Speaker 2: that it could be something that is worth investigating further. Doctor,

645
00:34:27,929 --> 00:34:28,979
Speaker 2: here's an assist

646
00:34:29,500 --> 00:34:34,968
Speaker 2: and that level of Jarvis like augmentation of what you

647
00:34:34,979 --> 00:34:40,638
Speaker 2: could potentially identify at scale through simply imaging and diagnostics

648
00:34:40,649 --> 00:34:46,810
Speaker 2: around health care outcomes starts to change. Um Your not

649
00:34:46,820 --> 00:34:49,120
Speaker 2: just the way of work but the way in which

650
00:34:49,129 --> 00:34:52,279
Speaker 2: you think about maybe even longevity and mortality and

651
00:34:52,540 --> 00:34:56,929
Speaker 2: the the delivery of health care in variety of different contexts.

652
00:34:57,010 --> 00:34:59,879
Speaker 2: These are the reasons why I'm just, you know,

653
00:35:00,179 --> 00:35:05,340
Speaker 2: much more optimistic than I am any other adjective. We

654
00:35:05,350 --> 00:35:09,520
Speaker 2: can't quite understand how some of our best partners. And

655
00:35:09,530 --> 00:35:13,280
Speaker 2: it's actually the biggest customers are thinking about using these

656
00:35:13,290 --> 00:35:17,419
Speaker 2: tools to come up with something that's really quite exciting.

657
00:35:17,540 --> 00:35:18,989
Speaker 2: And I'm going to go back to creation just for

658
00:35:19,000 --> 00:35:21,139
Speaker 2: a second because I think that the most exciting part

659
00:35:21,149 --> 00:35:23,040
Speaker 2: of around one of the most exciting parts about 8

660
00:35:23,050 --> 00:35:24,139
Speaker 2: to 10 years from now,

661
00:35:24,455 --> 00:35:28,745
Speaker 2: it's gonna be just a wide variety of content and

662
00:35:28,754 --> 00:35:31,395
Speaker 2: craft that we will be able to see. So there

663
00:35:31,405 --> 00:35:33,804
Speaker 2: was a Korean web lit, think of it like a

664
00:35:33,814 --> 00:35:36,554
Speaker 2: like a animated web drama. They used to put out

665
00:35:36,564 --> 00:35:40,485
Speaker 2: an episode every two weeks immediately after the, the launch

666
00:35:40,495 --> 00:35:41,924
Speaker 2: of gen A I, what they were able to do

667
00:35:41,935 --> 00:35:44,514
Speaker 2: is to use background shading and other types of tools

668
00:35:44,524 --> 00:35:47,495
Speaker 2: to go from an episode every two weeks to two

669
00:35:47,504 --> 00:35:48,594
Speaker 2: episodes per week.

670
00:35:49,010 --> 00:35:51,389
Speaker 2: So it's a four times throughput of the amount of

671
00:35:51,399 --> 00:35:53,790
Speaker 2: content that they could create. So you could be more

672
00:35:53,800 --> 00:35:56,750
Speaker 2: imaginative and focus more on the story, more on the narrative,

673
00:35:56,760 --> 00:35:57,350
Speaker 2: more on

674
00:35:57,780 --> 00:36:00,949
Speaker 2: maybe even the merchandizing. But the truth is that that

675
00:36:00,959 --> 00:36:04,139
Speaker 2: productivity gain, the forex just comes from basically doing things

676
00:36:04,149 --> 00:36:06,320
Speaker 2: that humans used to have to do in background shading

677
00:36:06,449 --> 00:36:09,399
Speaker 2: and now get delivered so that he releases the people

678
00:36:09,409 --> 00:36:13,120
Speaker 2: as it takes to build that story that that drama

679
00:36:13,379 --> 00:36:16,939
Speaker 2: into doing things that are perhaps even more creative. So

680
00:36:16,949 --> 00:36:19,850
Speaker 2: we will live in a in a horn of cleaning

681
00:36:19,860 --> 00:36:22,070
Speaker 2: when it comes to content, we'll have a variety of

682
00:36:22,080 --> 00:36:24,550
Speaker 2: different ways of thinking about um

683
00:36:25,129 --> 00:36:28,699
Speaker 2: how we might use this tool to create more. And

684
00:36:28,709 --> 00:36:31,780
Speaker 2: yet there's gonna be supples along the way where we

685
00:36:31,790 --> 00:36:34,189
Speaker 2: ask ourselves, wait, are we being inclusive? Is this fair

686
00:36:34,199 --> 00:36:36,569
Speaker 2: or is this, you know, we have the right accountability.

687
00:36:36,729 --> 00:36:38,429
Speaker 2: So I don't think that it's wrong for us to

688
00:36:38,439 --> 00:36:42,219
Speaker 2: constantly check to make sure that we're doing all the

689
00:36:42,229 --> 00:36:45,330
Speaker 2: right things but that the optimism I have for how

690
00:36:45,340 --> 00:36:48,969
Speaker 2: it might unlock in the real world is rooted in

691
00:36:48,979 --> 00:36:52,070
Speaker 2: things that we're already seeing in this year to upend.

692
00:36:52,570 --> 00:36:55,629
Speaker 1: Sure. Yeah, I, I don't think we have a problem

693
00:36:55,639 --> 00:36:58,070
Speaker 1: of too little as far as content is concerned. We

694
00:36:58,080 --> 00:37:00,270
Speaker 1: probably have a problem of plenty already, but I, I

695
00:37:00,280 --> 00:37:04,810
Speaker 1: hope that we get a quality quantity optimal mix uh

696
00:37:04,820 --> 00:37:08,589
Speaker 1: going forward. And while I fully appreciate that, you know,

697
00:37:08,600 --> 00:37:12,109
Speaker 1: companies like Microsoft are trying to instill without any nudge

698
00:37:12,120 --> 00:37:13,560
Speaker 1: necessarily from positive

699
00:37:14,110 --> 00:37:18,100
Speaker 1: a system of governance and internal uh sort of, you know,

700
00:37:18,439 --> 00:37:22,649
Speaker 1: cross checks. Uh It is true that, you know, through

701
00:37:22,659 --> 00:37:27,610
Speaker 1: waves of transformative technology, when the disruption takes place, regulators

702
00:37:27,620 --> 00:37:29,949
Speaker 1: and regulations sort of struggle to keep up with the

703
00:37:29,959 --> 00:37:34,009
Speaker 1: pace of change. And I think the last 20 years

704
00:37:34,020 --> 00:37:34,699
Speaker 1: with the internet

705
00:37:34,790 --> 00:37:38,679
Speaker 1: revolution to the social media proliferation, I think there have

706
00:37:38,689 --> 00:37:42,830
Speaker 1: been failings at times to protect consumer privacy, fake news,

707
00:37:42,840 --> 00:37:46,889
Speaker 1: election interference, et cetera. So how should we balance the

708
00:37:46,899 --> 00:37:50,050
Speaker 1: imperative of the business to go out and innovate and

709
00:37:50,060 --> 00:37:53,449
Speaker 1: come up with products for the customers and public interest

710
00:37:53,459 --> 00:37:55,810
Speaker 1: which are not necessarily always aligned?

711
00:37:56,810 --> 00:38:00,550
Speaker 2: Yeah, I think that you refer to some really big

712
00:38:00,560 --> 00:38:04,689
Speaker 2: disappointments and you know, speaking personally about what the internet

713
00:38:04,699 --> 00:38:08,350
Speaker 2: could have been. And uh I think that there is

714
00:38:08,590 --> 00:38:13,370
Speaker 2: uh an element of being thoughtful around safeguards that as

715
00:38:13,379 --> 00:38:16,750
Speaker 2: an industry plus the public sector we could have done

716
00:38:16,939 --> 00:38:21,010
Speaker 2: to make sure that we protect, um, Children or, or, uh,

717
00:38:21,020 --> 00:38:21,750
Speaker 2: identity

718
00:38:22,399 --> 00:38:25,979
Speaker 2: as we go forward. I think that working together on

719
00:38:25,989 --> 00:38:30,969
Speaker 2: a set of standards with the industry and governments in

720
00:38:30,979 --> 00:38:35,959
Speaker 2: lockstep is important. And there's this data that came from

721
00:38:35,969 --> 00:38:38,360
Speaker 2: Oliver Wyman. I'm looking at it right now. It was,

722
00:38:38,370 --> 00:38:41,759
Speaker 2: it was done with across 16 countries and I don't

723
00:38:41,770 --> 00:38:46,060
Speaker 2: mean it to refer to less regulation is good. But the,

724
00:38:46,070 --> 00:38:48,259
Speaker 2: if you look at the question of how often are

725
00:38:48,270 --> 00:38:50,860
Speaker 2: you using generative A I in your current job?

726
00:38:51,820 --> 00:38:56,279
Speaker 2: Three out of the top five are right here. India,

727
00:38:56,290 --> 00:39:00,969
Speaker 2: Indonesia and Singapore are 13 and five on that list

728
00:39:01,129 --> 00:39:05,419
Speaker 2: with UAE in Brazil, um being two and four. And

729
00:39:05,429 --> 00:39:06,899
Speaker 2: so you look at that, you say the, is it

730
00:39:06,909 --> 00:39:10,050
Speaker 2: the global South? Is it a light touch regulation or

731
00:39:10,060 --> 00:39:12,439
Speaker 2: is it optimism again? Is it an optimism that I

732
00:39:12,449 --> 00:39:14,939
Speaker 2: hear in India can basically improve my pro

733
00:39:15,044 --> 00:39:18,625
Speaker 2: activity by using a tool in Indonesia? I can basically

734
00:39:18,635 --> 00:39:23,195
Speaker 2: become Multilingual immediately by using tools or Singapore or UAE

735
00:39:23,205 --> 00:39:25,334
Speaker 2: leaning forward in their governments and saying we need to

736
00:39:25,344 --> 00:39:28,145
Speaker 2: experiment to think about ways in which we can embrace

737
00:39:28,155 --> 00:39:31,544
Speaker 2: this tool versus I hate to say it dead last

738
00:39:31,554 --> 00:39:34,054
Speaker 2: on this list. Do you have any guesses? Tw

739
00:39:34,514 --> 00:39:35,024
Speaker 1: don't

740
00:39:35,034 --> 00:39:35,824
Speaker 1: say us

741
00:39:35,975 --> 00:39:36,824
Speaker 2: Canada

742
00:39:37,324 --> 00:39:38,024
Speaker 1: really?

743
00:39:38,034 --> 00:39:43,185
Speaker 2: Us middle of the pack 46 Canada is the worst Australia.

744
00:39:43,195 --> 00:39:44,145
Speaker 2: Close second.

745
00:39:44,790 --> 00:39:49,030
Speaker 2: UK third. Why, why so much suspicion I asked my

746
00:39:49,040 --> 00:39:52,379
Speaker 2: Canadian relatives all the time. There's just this

747
00:39:53,010 --> 00:39:55,550
Speaker 2: worry about. Oh, you can't use that at school. You

748
00:39:55,560 --> 00:39:57,389
Speaker 2: can't just instantly. Do you use a calculator? Do you

749
00:39:57,399 --> 00:40:00,649
Speaker 2: think about research? Do you ask it some questions? I

750
00:40:00,659 --> 00:40:03,669
Speaker 2: have a Canadian friend. We, we, we both work for, um,

751
00:40:03,679 --> 00:40:06,959
Speaker 2: big tech companies now outside of Canada. But we asked

752
00:40:06,969 --> 00:40:09,020
Speaker 2: the question about something that we've been thinking about since

753
00:40:09,030 --> 00:40:11,620
Speaker 2: we were teenagers. And it's so interesting how general de

754
00:40:11,959 --> 00:40:15,550
Speaker 2: comes up with a better mathematical answer to that today

755
00:40:15,939 --> 00:40:18,580
Speaker 2: than when we were young. And we're using simulation and

756
00:40:18,590 --> 00:40:21,270
Speaker 2: other tools and just full disclosure, he said, Google

757
00:40:21,830 --> 00:40:24,060
Speaker 2: and he asked the question, do you think this is

758
00:40:24,070 --> 00:40:27,010
Speaker 2: 90 better than 99% of humans? I said, I'm not

759
00:40:27,020 --> 00:40:29,090
Speaker 2: sure it is yet. And then he said, you know

760
00:40:29,100 --> 00:40:30,729
Speaker 2: what it must be because it's better than you at

761
00:40:30,739 --> 00:40:33,340
Speaker 2: solving this problem. And I laughed and I said, yeah,

762
00:40:33,350 --> 00:40:35,770
Speaker 2: you're probably right. And so the problem

763
00:40:36,159 --> 00:40:38,020
Speaker 2: is something that you might see at work, it might

764
00:40:38,030 --> 00:40:39,379
Speaker 2: be something that you see in school, it might be

765
00:40:39,389 --> 00:40:41,350
Speaker 2: something you see in your research lab. But if you're

766
00:40:41,360 --> 00:40:44,929
Speaker 2: not using a tool that just hopefully helps you solve

767
00:40:44,939 --> 00:40:47,250
Speaker 2: that problem a little bit faster than what are you doing?

768
00:40:47,439 --> 00:40:49,879
Speaker 2: So you can't shut it off. You can't say no,

769
00:40:49,889 --> 00:40:51,889
Speaker 2: we're not going to have generated A A I, you

770
00:40:51,899 --> 00:40:54,109
Speaker 2: have to think about this tool in the right context

771
00:40:54,120 --> 00:40:57,280
Speaker 2: with the right safeguards in order to unlock the economy. So,

772
00:40:57,409 --> 00:40:59,260
Speaker 2: you know, I'm gonna get a lot of like flame

773
00:40:59,270 --> 00:41:02,500
Speaker 2: from friends in Canada. But we need to think about

774
00:41:02,510 --> 00:41:03,179
Speaker 2: why is it that

775
00:41:03,264 --> 00:41:06,395
Speaker 2: Indonesia are one and three? What is it, the UAE

776
00:41:06,405 --> 00:41:09,574
Speaker 2: is doing? That's gotten our attention and certainly Singapore as well.

777
00:41:09,584 --> 00:41:12,145
Speaker 2: And let's not forget, let me just count this all

778
00:41:12,155 --> 00:41:16,804
Speaker 2: China is uh eighth which is pretty good out of 16.

779
00:41:16,814 --> 00:41:19,625
Speaker 2: Um And, and there seems to be a cliff after

780
00:41:19,635 --> 00:41:22,455
Speaker 2: China where 60% of the people use it according to

781
00:41:22,465 --> 00:41:24,965
Speaker 2: this Oliver Wyman study. So not sure I fully answered

782
00:41:24,975 --> 00:41:27,844
Speaker 2: your question, but I say embrace it. Have safeguards, don't

783
00:41:27,854 --> 00:41:29,584
Speaker 2: be afraid of it because it does actually help you

784
00:41:29,594 --> 00:41:30,205
Speaker 2: do your job

785
00:41:31,080 --> 00:41:35,540
Speaker 1: fascinated by that survey. Very interesting. Um I want to

786
00:41:35,550 --> 00:41:38,600
Speaker 1: ask you the question. Maybe I could have started this

787
00:41:38,610 --> 00:41:42,229
Speaker 1: whole conversation with because these days, Gen A I open

788
00:41:42,300 --> 00:41:46,689
Speaker 1: A I and Microsoft are almost synonymous. So tell us

789
00:41:46,699 --> 00:41:49,860
Speaker 1: the depth and the breadth of the relationship and what's

790
00:41:49,870 --> 00:41:51,379
Speaker 1: next in this collaboration.

791
00:41:52,360 --> 00:41:55,060
Speaker 2: Yeah, I think the big takeaway here is that Microsoft

792
00:41:55,070 --> 00:41:59,949
Speaker 2: has multiple partners in generative A I. And uh we

793
00:41:59,959 --> 00:42:02,060
Speaker 2: have made um

794
00:42:03,050 --> 00:42:06,639
Speaker 2: partnerships with companies like Mistral in Europe. We saw G

795
00:42:06,649 --> 00:42:10,800
Speaker 2: 42 UAE, not just the language model companies but companies

796
00:42:10,810 --> 00:42:12,870
Speaker 2: that are doing great work on top of the model

797
00:42:12,879 --> 00:42:15,159
Speaker 2: as well. We made an investment in a company called

798
00:42:15,169 --> 00:42:18,139
Speaker 2: hidden layer that does cybersecurity to help protect your model,

799
00:42:18,300 --> 00:42:20,750
Speaker 2: to make sure that it is not assailed and corrupted

800
00:42:20,760 --> 00:42:24,750
Speaker 2: by uh bad actors who will try to taint what

801
00:42:24,760 --> 00:42:27,149
Speaker 2: it is that you tell people is the next best word.

802
00:42:27,520 --> 00:42:30,340
Speaker 2: But it does all start with our first um uh

803
00:42:30,350 --> 00:42:34,199
Speaker 2: relationship with open A I many years ago. Right. I mean,

804
00:42:34,209 --> 00:42:37,790
Speaker 2: it wasn't just a year and a half ago. Um

805
00:42:37,800 --> 00:42:41,939
Speaker 2: Our first um partnership with them was more, let's gather

806
00:42:41,949 --> 00:42:44,350
Speaker 2: a signal from this group of people that are uh

807
00:42:44,360 --> 00:42:47,360
Speaker 2: coming out of Y Combinator. And, and then Sam started

808
00:42:47,370 --> 00:42:49,989
Speaker 2: to run the company and things started to take off

809
00:42:50,000 --> 00:42:53,030
Speaker 2: with the Transformer model to the point where we had

810
00:42:53,040 --> 00:42:54,780
Speaker 2: a moment where we had to make a decision as

811
00:42:54,790 --> 00:42:56,540
Speaker 2: to whether or not we make that big bet.

812
00:42:56,685 --> 00:42:59,524
Speaker 2: And that was just before Chat G BT, we made

813
00:42:59,534 --> 00:43:02,425
Speaker 2: the big bet that big partnership with Open A I.

814
00:43:02,465 --> 00:43:04,714
Speaker 2: And then since then, we've made other bets which are

815
00:43:04,725 --> 00:43:07,935
Speaker 2: also very big around how we might think about generative

816
00:43:07,945 --> 00:43:11,844
Speaker 2: A I more broadly. But certainly the chat GP T

817
00:43:11,854 --> 00:43:15,635
Speaker 2: moment should not be underestimated for what it unlocked. And

818
00:43:15,645 --> 00:43:18,685
Speaker 2: we continue to be not just on their board but

819
00:43:18,695 --> 00:43:21,823
Speaker 2: also the beneficiary that every time you do a search

820
00:43:21,834 --> 00:43:25,205
Speaker 2: on uh or a query or uh prompt

821
00:43:25,610 --> 00:43:29,979
Speaker 2: uh or on open A I, it hits our servers

822
00:43:29,989 --> 00:43:32,330
Speaker 2: in the back end. So we obviously a beneficiary of

823
00:43:32,340 --> 00:43:35,070
Speaker 2: being the one that does the supercomputer. But that's the

824
00:43:35,080 --> 00:43:37,629
Speaker 2: case for so many other models as well so that

825
00:43:37,639 --> 00:43:40,739
Speaker 2: we can be in a not quite frenemy but in

826
00:43:40,750 --> 00:43:45,229
Speaker 2: a collaboration and cooperation's like competition with even

827
00:43:45,385 --> 00:43:49,604
Speaker 2: other models. And so it isn't perhaps as synonymous as

828
00:43:49,614 --> 00:43:54,604
Speaker 2: you might say. But um we are clearly uh AAA

829
00:43:54,614 --> 00:43:58,645
Speaker 2: strong partner of theirs and a beneficiary of when they innovate,

830
00:43:58,905 --> 00:44:02,364
Speaker 2: we benefit and everyone benefits. But there are other sources

831
00:44:02,375 --> 00:44:06,364
Speaker 2: of innovation uh back to, you know, it's just unevenly

832
00:44:06,375 --> 00:44:07,125
Speaker 2: distributed

833
00:44:08,659 --> 00:44:11,719
Speaker 1: a year from now. Will you and I be able

834
00:44:11,729 --> 00:44:15,560
Speaker 1: to do a podcast in English and have it simultaneously

835
00:44:15,600 --> 00:44:17,389
Speaker 1: capacitated in Bengali and

836
00:44:17,399 --> 00:44:17,830
Speaker 2: it already

837
00:44:17,840 --> 00:44:20,719
Speaker 2: is in place. So we invested in Saran, which is

838
00:44:20,729 --> 00:44:22,360
Speaker 2: a company that um

839
00:44:23,010 --> 00:44:25,979
Speaker 2: we put some thought into Sara. I'll be a little

840
00:44:25,989 --> 00:44:29,299
Speaker 2: bit more clear. Uh and some resources into a company

841
00:44:29,310 --> 00:44:33,439
Speaker 2: that basically helps us translate into 19 indic languages, I

842
00:44:33,449 --> 00:44:40,139
Speaker 2: think right now, indie Bengali Urdu uh are probably already

843
00:44:40,149 --> 00:44:44,469
Speaker 2: covered by our live teams app. So this is on Zoom,

844
00:44:44,479 --> 00:44:46,270
Speaker 2: but it had it been on teams. It would have

845
00:44:46,280 --> 00:44:51,138
Speaker 2: been real time Bengali Bangla uh translation with voice if

846
00:44:51,149 --> 00:44:52,219
Speaker 2: we wanted it to be

847
00:44:52,510 --> 00:44:55,760
Speaker 2: um I use this tool perhaps more than any other tool,

848
00:44:55,840 --> 00:44:59,279
Speaker 2: just the ability to say. What did I miss? Uh

849
00:44:59,290 --> 00:45:02,850
Speaker 2: when was the name Zia mentioned? Did anyone tell a joke?

850
00:45:03,530 --> 00:45:06,310
Speaker 2: Did anyone tell a joke about Zia and then like,

851
00:45:06,320 --> 00:45:08,739
Speaker 2: it's just so interesting to think about how we can

852
00:45:08,750 --> 00:45:13,429
Speaker 2: very quickly query in, um, what has been said and, and,

853
00:45:13,439 --> 00:45:18,030
Speaker 2: and understand it independent of language now. So we're already there. Um,

854
00:45:18,040 --> 00:45:20,009
Speaker 2: if the, if you say what's the one year from

855
00:45:20,020 --> 00:45:21,709
Speaker 2: now thing that you think might be unlocked that we

856
00:45:21,719 --> 00:45:24,429
Speaker 2: won't be able to do? I'm gonna go back to teaching.

857
00:45:24,800 --> 00:45:28,560
Speaker 2: I think there's a possibility that something that our 11

858
00:45:28,570 --> 00:45:30,949
Speaker 2: or 14 year old learn

859
00:45:31,629 --> 00:45:36,010
Speaker 2: will through the tool will far exceed

860
00:45:36,860 --> 00:45:39,270
Speaker 2: what it took us many, many years of education and

861
00:45:39,280 --> 00:45:42,120
Speaker 2: living on this planet to get good at. And we

862
00:45:42,129 --> 00:45:44,669
Speaker 2: will be shocked at how quickly

863
00:45:45,389 --> 00:45:48,850
Speaker 2: our kids are able to do specific things for which

864
00:45:48,860 --> 00:45:51,350
Speaker 2: we have absolutely no clue how they learn how to

865
00:45:51,360 --> 00:45:53,280
Speaker 2: do that. It is an element of magic to all

866
00:45:53,290 --> 00:45:58,129
Speaker 2: great innovation, right? And uh the ability to use this

867
00:45:58,139 --> 00:46:00,419
Speaker 2: tool to do so many new interesting things that I

868
00:46:00,429 --> 00:46:01,320
Speaker 2: think is unwritten.

869
00:46:02,179 --> 00:46:07,340
Speaker 1: I am waiting for that infinitely patient and infinitely empathetic instructor.

870
00:46:07,350 --> 00:46:09,069
Speaker 1: I'm really, really waiting for, not just for my son

871
00:46:09,080 --> 00:46:10,590
Speaker 1: but for myself as well. I'm sure there are a

872
00:46:10,600 --> 00:46:13,320
Speaker 1: bunch of things that I could learn. Uh And I

873
00:46:13,330 --> 00:46:16,319
Speaker 1: would need a very patient teacher for that. Yeah, I

874
00:46:16,330 --> 00:46:18,709
Speaker 1: so much appreciate you coming on the show and this

875
00:46:18,719 --> 00:46:21,750
Speaker 1: is a very, very valuable insightful conversation. Thank you so

876
00:46:21,760 --> 00:46:22,919
Speaker 1: much for your time and insights.

877
00:46:23,389 --> 00:46:26,159
Speaker 2: Thank you so much for having me. I really appreciate it.

878
00:46:26,169 --> 00:46:28,479
Speaker 2: It's long overdue. Thanks again. Thank

879
00:46:28,489 --> 00:46:31,360
Speaker 1: you and thanks to our listeners and viewers as well.

880
00:46:31,370 --> 00:46:34,750
Speaker 1: Kobe Time was produced by Ken Delbridge at Spy studios,

881
00:46:34,820 --> 00:46:38,090
Speaker 1: Violet Le and Daisy Sherma provided additional assistance. It is

882
00:46:38,100 --> 00:46:39,610
Speaker 1: for information only and design

883
00:46:39,685 --> 00:46:44,354
Speaker 1: represent any investment advice. All 126 episodes of Cookie Time

884
00:46:44,364 --> 00:46:47,945
Speaker 1: are available on Apple, Google and Spotify, as well as

885
00:46:47,955 --> 00:46:51,215
Speaker 1: youtube as for our research publications, webinars and live streams.

886
00:46:51,225 --> 00:46:54,465
Speaker 1: You can find them all by Googling D BS research Library.

887
00:46:54,675 --> 00:46:55,864
Speaker 1: Have a great day.