1 00:00:00,080 --> 00:00:02,797 Think about it. The the way I come at 2 00:00:02,880 --> 00:00:06,797 this is not that um this has always been 3 00:00:06,880 --> 00:00:11,437 the arc of computation, right? You can sort of take it in the last 30 years or 4 00:00:11,520 --> 00:00:13,597 the last 70 years. It's always been 5 00:00:13,680 --> 00:00:16,077 about can you digitize 6 00:00:16,160 --> 00:00:19,917 artifacts on about people, places, and 7 00:00:20,000 --> 00:00:24,157 things and then build analytical and 8 00:00:24,240 --> 00:00:25,517 predictive power. [clears throat] Right? 9 00:00:25,600 --> 00:00:30,237 That's what the mainframes did. That's what the mini computer did. That's what 10 00:00:30,320 --> 00:00:34,317 the client server error did. That's what the web error did. The mobile cloud 11 00:00:34,400 --> 00:00:36,557 error did. So it like it depend 12 00:00:36,640 --> 00:00:39,677 irrespective of which paradigm or 13 00:00:39,760 --> 00:00:43,677 platform. It has been one continuous arc 14 00:00:43,760 --> 00:00:46,957 of saying let's make better sense of 15 00:00:47,040 --> 00:00:48,877 this world 16 00:00:48,960 --> 00:00:53,597 um by reasoning about it in digital form 17 00:00:53,680 --> 00:00:58,477 because in some sense once you have these artifacts in digital form you can 18 00:00:58,560 --> 00:01:01,277 use a more malleable resource like software 19 00:01:01,360 --> 00:01:01,997 >> right 20 00:01:02,080 --> 00:01:05,917 >> um which doesn't have the same type of 21 00:01:06,000 --> 00:01:08,877 you know I'll call it u marginal cost 22 00:01:08,960 --> 00:01:11,437 economics associated ated with it that 23 00:01:11,520 --> 00:01:14,317 allows us to then build more insight and 24 00:01:14,400 --> 00:01:17,597 more uh more capability and in that 25 00:01:17,680 --> 00:01:20,157 context AI I would say is of the same 26 00:01:20,240 --> 00:01:22,717 class at least like the web or the 27 00:01:22,800 --> 00:01:26,477 internet um or mobile or PC or the cloud 28 00:01:26,560 --> 00:01:30,637 or and maybe even greater and so to me 29 00:01:30,720 --> 00:01:33,917 right now where we are is you know let's 30 00:01:34,000 --> 00:01:36,157 take just what's happened with software 31 00:01:36,240 --> 00:01:38,717 engineering right which is one you know 32 00:01:38,800 --> 00:01:41,517 is knowledge work um you know you could 33 00:01:41,600 --> 00:01:43,837 say it's elite knowledge work 34 00:01:43,920 --> 00:01:46,797 >> it started off uh you know in fact my 35 00:01:46,880 --> 00:01:51,597 own belief in this generation of AI and its capability 36 00:01:51,680 --> 00:01:55,917 uh really got built up when I first saw GitHub copilot do code completions right 37 00:01:56,000 --> 00:02:00,477 so for the longest time we had the dream that if you're a software developer can 38 00:02:00,560 --> 00:02:02,877 you predict the next sort of word or the 39 00:02:02,960 --> 00:02:06,557 next uh line of code uh and suddenly it 40 00:02:06,640 --> 00:02:09,197 started working with these models u then 41 00:02:09,280 --> 00:02:11,197 You said okay if I can do that then can 42 00:02:11,280 --> 00:02:14,397 I actually go and bring back you know 43 00:02:14,480 --> 00:02:18,717 the flow for a software developer by going to a chat session and asking any 44 00:02:18,800 --> 00:02:23,117 question and it comes back with answers that then uh you can use in your coding 45 00:02:23,200 --> 00:02:26,797 flow right so that was the next thing then you said well if that's working can 46 00:02:26,880 --> 00:02:31,917 I assign it small tasks that was the agent mode uh now you have complete 47 00:02:32,000 --> 00:02:36,477 autonomous agents where you can give it your entire project right it can work uh 48 00:02:36,560 --> 00:02:39,597 you know >> 24/7 >> it can work for 24/7 7. I mean it's 49 00:02:39,680 --> 00:02:43,597 still we've got some ways to go for these things to remain coherent long 50 00:02:43,680 --> 00:02:46,077 time but nevertheless it's getting 51 00:02:46,160 --> 00:02:48,717 better and better and interestingly 52 00:02:48,800 --> 00:02:51,517 enough you look at it uh the software 53 00:02:51,600 --> 00:02:55,997 developer still is got a lot of agency in it right so that's why I kind of 54 00:02:56,080 --> 00:02:59,677 still think that you know going and thinking of these as somehow living 55 00:02:59,760 --> 00:03:03,597 outside of the realm of human agency is probably not the right way to think 56 00:03:03,680 --> 00:03:06,077 about it in fact the way to perhaps 57 00:03:06,160 --> 00:03:11,197 conceive it like it's let's say in early 80s If somebody had come to us and said 58 00:03:11,280 --> 00:03:15,757 what 4 billion people are going to wake up every morning and start typing you 59 00:03:15,840 --> 00:03:20,637 would have said why right you know we have a like we have a typist pool that's 60 00:03:20,720 --> 00:03:24,397 good enough we don't need 4 billion people but we that's what happened right 61 00:03:24,480 --> 00:03:27,277 we invented this entire class of thing 62 00:03:27,360 --> 00:03:29,357 called knowledge work where people 63 00:03:29,440 --> 00:03:33,037 started really using computers uh to go 64 00:03:33,120 --> 00:03:38,477 amplify what we were trying to achieve uh using software I think in the context 65 00:03:38,560 --> 00:03:40,877 of AI that same thing is going to happen. 66 00:03:40,960 --> 00:03:43,917 >> Um it's not like you know what is 67 00:03:44,000 --> 00:03:48,077 hardcore coding is going to remain hardcore coding forever. It's just that 68 00:03:48,160 --> 00:03:49,597 the levels of abstraction are going to 69 00:03:49,680 --> 00:03:52,637 change. Uh but we also are going to have 70 00:03:52,720 --> 00:03:55,117 code as output just like documents. In 71 00:03:55,200 --> 00:03:59,917 fact, one of Bill's things at Microsoft from the day I joined in '92 always was 72 00:04:00,000 --> 00:04:01,597 what's the real difference between a 73 00:04:01,680 --> 00:04:04,557 document, a website, and an application, 74 00:04:04,640 --> 00:04:07,277 right? It's the lack of sort of software 75 00:04:07,360 --> 00:04:09,917 that can transform itself. Interestingly 76 00:04:10,000 --> 00:04:11,997 enough, AI finally gives us that, right? 77 00:04:12,080 --> 00:04:15,837 Which is I can write a document. I can just say no, I don't want it as a 78 00:04:15,920 --> 00:04:19,837 document. I want it as a website. It'll just transform that document using code 79 00:04:19,920 --> 00:04:22,237 into a website. I say I don't like the 80 00:04:22,320 --> 00:04:23,997 website. I want an app. It'll write more 81 00:04:24,080 --> 00:04:27,197 code to transform it. So that reasoning 82 00:04:27,280 --> 00:04:29,757 and cap reasoning capabilities that 83 00:04:29,840 --> 00:04:32,077 prediction capabilities that ability to 84 00:04:32,160 --> 00:04:35,517 take action remain long-term coherent is 85 00:04:35,600 --> 00:04:40,477 all improving. Um and our job though is 86 00:04:40,560 --> 00:04:45,037 to parlay this like take even what you at BlackRock are doing right when you're 87 00:04:45,120 --> 00:04:49,197 bring taking something like say co-pilot plus Aladdin and bringing those things 88 00:04:49,280 --> 00:04:52,797 together [clears throat] >> to improve the productivity in the firm 89 00:04:52,880 --> 00:04:56,317 for the decisions you want to make right with your data 90 00:04:56,400 --> 00:04:58,237 >> I could just tell you from at our firm 91 00:04:58,320 --> 00:05:00,557 things that would take 12 hours to 92 00:05:00,640 --> 00:05:06,077 compute now takes minutes for us processing $14 trillion of other 93 00:05:06,160 --> 00:05:11,357 people's money with hundreds of thousands of different um mandates um we 94 00:05:11,440 --> 00:05:14,717 could do that instantaneous and we you know that to me if it wasn't for the 95 00:05:14,800 --> 00:05:18,877 technology and AI today we would not be able to function to the scale that we're 96 00:05:18,960 --> 00:05:22,717 operating >> that's right and so to me that one firm 97 00:05:22,800 --> 00:05:25,357 at a time one country at a time if we 98 00:05:25,440 --> 00:05:28,477 can really take these tokens and bend 99 00:05:28,560 --> 00:05:32,477 the curve of productivity then there is surplus everywhere and that's really the 100 00:05:32,560 --> 00:05:35,517 goal >> well surplus could be scary too does 101 00:05:35,600 --> 00:05:38,797 does surplus surplus mean fewer workers? 102 00:05:38,880 --> 00:05:43,037 What do we mean by surplus? And so, you know, the I'm going to tie that into my 103 00:05:43,120 --> 00:05:45,277 second question about AI diffusion. 104 00:05:45,360 --> 00:05:45,997 >> Yeah. 105 00:05:46,080 --> 00:05:48,877 >> To me, the the whole realization of AI 106 00:05:48,960 --> 00:05:52,317 for any society and also for a more 107 00:05:52,400 --> 00:05:55,117 balanced world is making sure that it's 108 00:05:55,200 --> 00:05:57,837 diffused and accessible and available 109 00:05:57,920 --> 00:06:00,157 across the world. So what you know can 110 00:06:00,240 --> 00:06:02,797 you describe how this process of 111 00:06:02,880 --> 00:06:07,997 diffusion uh across economies across companies across people and countries 112 00:06:08,080 --> 00:06:11,997 how does that play out? >> Yeah I think 113 00:06:12,080 --> 00:06:16,317 that this is the real question right because one of the uh things right now 114 00:06:16,400 --> 00:06:18,957 the zeitgeist is a little bit about the 115 00:06:19,040 --> 00:06:22,397 admiration for AI in its abstract form 116 00:06:22,480 --> 00:06:26,397 or as as technology. [clears throat] uh 117 00:06:26,480 --> 00:06:28,637 but I think we as even a global 118 00:06:28,720 --> 00:06:31,517 community um have to get to a point 119 00:06:31,600 --> 00:06:34,477 where we are using this to do something 120 00:06:34,560 --> 00:06:38,157 useful uh that changes uh the outcomes 121 00:06:38,240 --> 00:06:42,397 of people and communities and countries and industries right otherwise I don't 122 00:06:42,480 --> 00:06:44,877 think uh this makes much sense right in 123 00:06:44,960 --> 00:06:47,197 fact I would say we will quickly lose 124 00:06:47,280 --> 00:06:49,837 even the social permission 125 00:06:49,920 --> 00:06:52,477 uh to actually take something like uh 126 00:06:52,560 --> 00:06:56,877 energy which is a scarce resource and use it to generate these tokens if these 127 00:06:56,960 --> 00:07:00,637 tokens are not improving health outcomes, education outcomes, public 128 00:07:00,720 --> 00:07:05,037 sector efficiency, private sector competitiveness across all sectors, 129 00:07:05,120 --> 00:07:07,597 small and large, right? And that to me 130 00:07:07,680 --> 00:07:10,637 is ultimately the goal. So therefore I 131 00:07:10,720 --> 00:07:13,677 think really diffusion is everything. 132 00:07:13,760 --> 00:07:16,237 And so the way it happens is let's sort 133 00:07:16,320 --> 00:07:19,997 of unpack this uh on the supply side 134 00:07:20,080 --> 00:07:22,157 what needs to happen in each country is 135 00:07:22,240 --> 00:07:26,077 the tokens per dollar per watt have to 136 00:07:26,160 --> 00:07:31,117 sort of monotonically get more efficient and better right so to some degree even 137 00:07:31,200 --> 00:07:34,717 what we're trying to do with the investments the two firms are doing uh 138 00:07:34,800 --> 00:07:39,117 around the world is to just say that like let's make sure that the supply is 139 00:07:39,200 --> 00:07:41,597 there which is uh everything from the 140 00:07:41,680 --> 00:07:46,797 chips on down ultimately to these token factories that get deployed everywhere. 141 00:07:46,880 --> 00:07:50,717 By the way, there's not one token factory. This token factory is the first 142 00:07:50,800 --> 00:07:54,157 thing that's going to be diffused all around the world. It's just like 143 00:07:54,240 --> 00:07:57,117 electricity, right? You just need a 144 00:07:57,200 --> 00:08:01,917 ubiquitous grid of uh energy and tokens 145 00:08:02,000 --> 00:08:04,477 that then will power the rest of the 146 00:08:04,560 --> 00:08:06,157 economy. Right? So that's I think one 147 00:08:06,240 --> 00:08:09,357 side of it. Then the demand side of this 148 00:08:09,440 --> 00:08:12,397 is a little bit like every firm has to 149 00:08:12,480 --> 00:08:15,677 start by using it. If I look back even 150 00:08:15,760 --> 00:08:18,717 you know when the PCs first came out or 151 00:08:18,800 --> 00:08:22,957 the personal computing era started I I loved you know I think Jobs had a nice 152 00:08:23,040 --> 00:08:27,837 metaphor he called it the bicycle for the mind. Uh Bill had a metaphor which I 153 00:08:27,920 --> 00:08:31,037 remember was like information at your fingertips, right? These two metaphors 154 00:08:31,120 --> 00:08:34,957 were great like which allowed us to say that's what it is. It's a tool that I 155 00:08:35,040 --> 00:08:36,957 will use to get information at my 156 00:08:37,040 --> 00:08:38,957 fingertips. I'll use it as a cognitive 157 00:08:39,040 --> 00:08:42,797 amplifier. Now I think that's what we 158 00:08:42,880 --> 00:08:45,997 have. But you know 10x 100x right? So in 159 00:08:46,080 --> 00:08:48,877 some sense you as a as every knowledge 160 00:08:48,960 --> 00:08:52,077 worker you now have access to infinite 161 00:08:52,160 --> 00:08:54,077 minds. That's the way I think about it. 162 00:08:54,160 --> 00:08:58,077 Right? So there's a Turing award winner uh called Raj Ready who had this nice 163 00:08:58,160 --> 00:09:02,717 metaphor of AI and he had this long before uh even generative AI he said 164 00:09:02,800 --> 00:09:05,037 either either it's a cognitive amplifier 165 00:09:05,120 --> 00:09:07,757 or it's a guardian angel right so if you 166 00:09:07,840 --> 00:09:11,517 think of AI as that um then that in the 167 00:09:11,600 --> 00:09:15,677 global workforce right when a doctor can 168 00:09:15,760 --> 00:09:20,077 get to a patient spend more time with the patient because the AI is doing the 169 00:09:20,160 --> 00:09:22,317 transcription and entering the records 170 00:09:22,400 --> 00:09:25,197 in the EMR system entering the right 171 00:09:25,280 --> 00:09:27,917 billing code so that the health care uh 172 00:09:28,000 --> 00:09:30,077 you know industry is better served 173 00:09:30,160 --> 00:09:33,117 across the payer the provider and uh and 174 00:09:33,200 --> 00:09:35,997 the patient ultimately right that's an 175 00:09:36,080 --> 00:09:38,077 an outcome that I think all of us can 176 00:09:38,160 --> 00:09:41,037 benefit from so I feel ultimately it's 177 00:09:41,120 --> 00:09:43,597 going to require real leadership on the 178 00:09:43,680 --> 00:09:46,157 private sector and the public sector to 179 00:09:46,240 --> 00:09:48,957 ensure the diffusion happens and the one 180 00:09:49,040 --> 00:09:52,077 thing I'll mention Larry is skilling 181 00:09:52,160 --> 00:09:56,077 right so In some sense the thing that 182 00:09:56,160 --> 00:09:59,837 diffusion is very strongly correlated to 183 00:09:59,920 --> 00:10:02,797 one thing alone which is how broadly are 184 00:10:02,880 --> 00:10:07,037 people skilled in using this. Um and 185 00:10:07,120 --> 00:10:09,277 interestingly enough, I think if mobile 186 00:10:09,360 --> 00:10:12,637 has taught us one thing is it it's 187 00:10:12,720 --> 00:10:16,637 actually distinct from what happened in the PC, right? Uh I remember even 188 00:10:16,720 --> 00:10:18,957 growing up in the global south, uh there 189 00:10:19,040 --> 00:10:21,757 used to be a real relationship between 190 00:10:21,840 --> 00:10:25,117 learning Excel skills or word skills and 191 00:10:25,200 --> 00:10:28,717 getting a job. Um you know, right now, 192 00:10:28,800 --> 00:10:33,277 um what's the model in in mobile? It's kind of created the same opportunity, 193 00:10:33,360 --> 00:10:35,677 but it's been a lot more consumptionled. 194 00:10:35,760 --> 00:10:40,077 It's these creator economy and what have you, but it has not been about sort of, 195 00:10:40,160 --> 00:10:43,677 oh wow, here's how you get a healthcare job or here is how you get a finance job 196 00:10:43,760 --> 00:10:45,837 or here is how you get, you know, you 197 00:10:45,920 --> 00:10:48,477 get ahead professionally. 198 00:10:48,560 --> 00:10:51,597 Um, and that needs to come back, right? 199 00:10:51,680 --> 00:10:53,037 People need to say, oh, I pick up this 200 00:10:53,120 --> 00:10:56,077 AI skill and now I'm a better provider 201 00:10:56,160 --> 00:10:58,877 of some product or service in the real 202 00:10:58,960 --> 00:11:00,957 economy. So it's [cough] it's very easy 203 00:11:01,040 --> 00:11:03,757 to see how mobile and the diffusion of 204 00:11:03,840 --> 00:11:06,077 mobile how it transformed economies 205 00:11:06,160 --> 00:11:09,117 especially in the global south. How does 206 00:11:09,200 --> 00:11:11,597 how does this you know to me I I I just 207 00:11:11,680 --> 00:11:14,637 read a research report that said um the 208 00:11:14,720 --> 00:11:17,597 applications for AI so far are heavily 209 00:11:17,680 --> 00:11:21,837 weighted towards those who are educated or educated economies. 210 00:11:21,920 --> 00:11:26,957 And so does that create that you know more of a bifurcation more polarization? 211 00:11:27,040 --> 00:11:31,517 How do we ensure that that that diffusion is spread evenly? How do we 212 00:11:31,600 --> 00:11:33,117 make sure that we're not [clears throat] 213 00:11:33,200 --> 00:11:39,037 leaving major portions of society or the world behind because I think that's 214 00:11:39,120 --> 00:11:42,077 that's going to [clears throat] be the big issue for us going forward. Yeah. So 215 00:11:42,160 --> 00:11:43,677 it's it's it's interesting right this is 216 00:11:43,760 --> 00:11:46,797 one of those times when 217 00:11:46,880 --> 00:11:50,317 uh by definition and because of the 218 00:11:50,400 --> 00:11:54,317 rails that have been established you know as you said right which is 219 00:11:54,400 --> 00:11:57,197 >> uh what's happened with mobile as well 220 00:11:57,280 --> 00:12:01,677 as what's happened with um uh you know essentially connectivity 221 00:12:01,760 --> 00:12:05,277 >> right >> you have the ability to sort of deliver 222 00:12:05,360 --> 00:12:08,637 the tokens pretty evenly around the world >> right 223 00:12:08,720 --> 00:12:13,597 >> a lot more so than let's say uh the PC C era or even the beginning of the mobile 224 00:12:13,680 --> 00:12:15,277 era, right? Because it took a long time 225 00:12:15,360 --> 00:12:17,597 for even the mo smartphone in particular 226 00:12:17,680 --> 00:12:21,197 to penetrate um all of the world whereas 227 00:12:21,280 --> 00:12:23,597 now it's not the case, right? These 228 00:12:23,680 --> 00:12:27,037 models and their outputs are pretty much 229 00:12:27,120 --> 00:12:28,797 available everywhere. And so the 230 00:12:28,880 --> 00:12:34,237 question to me is what's the use cases that make sense, 231 00:12:34,320 --> 00:12:37,677 right? It's one, in fact, one of the demos I always go back to. I think this 232 00:12:37,760 --> 00:12:40,157 was even in the beginning of 23 was a 233 00:12:40,240 --> 00:12:43,037 rural Indian farmer was able to use a 234 00:12:43,120 --> 00:12:46,317 bot built on I think a very early GPT3 235 00:12:46,400 --> 00:12:50,317 or 25 even uh essentially to reason over 236 00:12:50,400 --> 00:12:52,477 some farm subsidies that he had heard 237 00:12:52,560 --> 00:12:54,957 about in a local language and had it 238 00:12:55,040 --> 00:12:58,557 even in that very early days uh have it 239 00:12:58,640 --> 00:13:00,557 even show some agentic behavior, right? 240 00:13:00,640 --> 00:13:02,477 Like go complete a form for me. So in 241 00:13:02,560 --> 00:13:04,477 some sense it took uh you know it 242 00:13:04,560 --> 00:13:09,037 brought back agency to someone uh who 243 00:13:09,120 --> 00:13:13,197 perhaps didn't have that because the technology was so much more accessible. 244 00:13:13,280 --> 00:13:15,997 So I I do think it's in our hands even 245 00:13:16,080 --> 00:13:19,197 in the global south to use it uh to 246 00:13:19,280 --> 00:13:21,757 create I would say more of that 247 00:13:21,840 --> 00:13:26,887 opportunity where there isn't one. Um but I think what the necessary 248 00:13:26,970 --> 00:13:30,637 [clears throat] conditions still are do you have uh the 249 00:13:30,720 --> 00:13:35,277 capital investment being put in uh do you even have an environment for capital 250 00:13:35,360 --> 00:13:39,277 because in an interesting way we are for example as hyperscalers investing all 251 00:13:39,360 --> 00:13:43,517 over right including the global south uh so as long as there's an environment 252 00:13:43,600 --> 00:13:46,477 which attracts the capital investment >> and you see the demand 253 00:13:46,560 --> 00:13:49,437 >> and you see and then yeah the demand is there yeah 254 00:13:49,520 --> 00:13:53,677 >> um and so the question is how do you have a set of policies that allow for 255 00:13:53,760 --> 00:13:57,997 both the capital to come in for it to find nexus where there are certain 256 00:13:58,080 --> 00:14:00,237 things by the way private capital can do 257 00:14:00,320 --> 00:14:02,157 certain things that public capital only 258 00:14:02,240 --> 00:14:05,197 can do for example the grid right it's 259 00:14:05,280 --> 00:14:10,317 not I mean grid in most countries is sort of fundamentally driven by governments 260 00:14:10,400 --> 00:14:12,317 >> public >> and public and so if you [clears throat] 261 00:14:12,400 --> 00:14:14,637 don't have a sophisticated sort or 262 00:14:14,720 --> 00:14:19,517 rather if you don't have a a real approach to modernizing the grid uh that 263 00:14:19,600 --> 00:14:22,957 will hold things back I mean there's a lot of talk about behind the meter and 264 00:14:23,040 --> 00:14:25,677 so on and Yes, there's some amount of that we can do ourselves. 265 00:14:25,760 --> 00:14:27,837 >> We can do that in the US. Many countries can't. 266 00:14:27,920 --> 00:14:29,437 >> Exactly. And it's not long-term 267 00:14:29,520 --> 00:14:31,357 scalable, right? I mean, like to me, a 268 00:14:31,440 --> 00:14:34,877 long-term scalable solution is to have 269 00:14:34,960 --> 00:14:39,677 uh you know, all of these token factories, part of the real economy, 270 00:14:39,760 --> 00:14:44,317 connected to the grid, connected to the telco network, delivering uh just like 271 00:14:44,400 --> 00:14:46,317 we delivered bits, you have to deliver 272 00:14:46,400 --> 00:14:49,357 tokens plus bits. Um and that's kind of 273 00:14:49,440 --> 00:14:53,757 what's going to drive an app scale whether it's in the global south or in 274 00:14:53,840 --> 00:14:57,037 uh on the developed world. >> So so many people talk about there may 275 00:14:57,120 --> 00:14:59,037 be an AI bubble. I mean the most 276 00:14:59,120 --> 00:15:00,397 important thing that we see as an 277 00:15:00,480 --> 00:15:04,637 investor is the the democratization of 278 00:15:04,720 --> 00:15:07,437 technology is and the diffusion of that 279 00:15:07,520 --> 00:15:09,997 technology really does then transform 280 00:15:10,080 --> 00:15:15,117 the demand and the the companies or the countries that diffuse it fastest are 281 00:15:15,200 --> 00:15:16,637 going to be the ultimate winners. about 282 00:15:16,720 --> 00:15:18,477 the technology creator. 283 00:15:18,560 --> 00:15:22,717 >> That's that's you know it's it's it's 284 00:15:22,800 --> 00:15:25,117 for this not to be a bubble 285 00:15:25,200 --> 00:15:25,677 >> y 286 00:15:25,760 --> 00:15:29,917 >> by definition it requires that the 287 00:15:30,000 --> 00:15:32,157 benefits of this are much more evenly 288 00:15:32,240 --> 00:15:35,357 spread. I mean I think a tellt tell sign 289 00:15:35,440 --> 00:15:38,717 of if it's a bubble would be if all 290 00:15:38,800 --> 00:15:41,917 we're talking about are the tech firms. 291 00:15:42,000 --> 00:15:45,037 Uh right? If uh all we talk about is 292 00:15:45,120 --> 00:15:49,037 what's happening to the technology side that then that's by day you know it's 293 00:15:49,120 --> 00:15:53,437 just purely supply side >> right >> uh ultimately if we are not talking 294 00:15:53,520 --> 00:15:57,597 about wow here is a drug comp you know drug that was sort of uh brought into 295 00:15:57,680 --> 00:16:01,757 the market that's super successful because it was uh AI accelerated the 296 00:16:01,840 --> 00:16:03,757 clinical trial it's not even the magical 297 00:16:03,840 --> 00:16:05,757 molecule right it's kind of even the 298 00:16:05,840 --> 00:16:08,877 rest of what is uh needed in order to 299 00:16:08,960 --> 00:16:11,517 make something much more relevant right 300 00:16:11,600 --> 00:16:16,717 um and So the more [clears throat] we have uh and by the way it's happening 301 00:16:16,800 --> 00:16:18,397 right. So I'm not sort of saying that 302 00:16:18,480 --> 00:16:21,277 that's why I'm much more confident that 303 00:16:21,360 --> 00:16:23,757 this is a technology that will in fact 304 00:16:23,840 --> 00:16:27,117 build on the rails of cloud and mobile 305 00:16:27,200 --> 00:16:30,287 diffuse faster and bend the productivity 306 00:16:30,370 --> 00:16:34,397 [clears throat] curve and bring local surplus and 307 00:16:34,480 --> 00:16:36,957 economic growth all around the world. 308 00:16:37,040 --> 00:16:41,277 Not just economic growth driven by capital expenses. 309 00:16:41,360 --> 00:16:46,557 uh right because that's it's a narrow point in time uh calculation. right now 310 00:16:46,640 --> 00:16:48,797 that's what we're seeing more >> that's what we're seeing in you know you 311 00:16:48,880 --> 00:16:50,317 know in in the developed world in 312 00:16:50,400 --> 00:16:53,917 particular uh but remember my capital 313 00:16:54,000 --> 00:16:57,437 like that the one thing that you know is definitely we spending a lot of it in 314 00:16:57,520 --> 00:17:00,797 the United States but 50% of it is also all over the world 315 00:17:00,880 --> 00:17:04,637 >> right >> um and so interesting enough it depends 316 00:17:04,720 --> 00:17:09,037 on uh demand all over the world and the demand all over the world will only be 317 00:17:09,120 --> 00:17:13,357 there if there is local surplus all over the world and so that's sort of the way 318 00:17:13,440 --> 00:17:17,357 I see the equation >> so let's drill down a little more. As AI 319 00:17:17,440 --> 00:17:21,357 diffuses, obviously organizations, companies, 320 00:17:21,440 --> 00:17:23,757 governments are going to have to evolve. 321 00:17:23,840 --> 00:17:27,117 Now getting to the demand side. So how 322 00:17:27,200 --> 00:17:28,477 do you think the structure of 323 00:17:28,560 --> 00:17:32,397 organizations is changing in an AI world 324 00:17:32,480 --> 00:17:35,837 across roles, across teams, management? 325 00:17:35,920 --> 00:17:39,357 I'm I'm sure um Microsoft has evolved 326 00:17:39,440 --> 00:17:42,237 itself. So it probably be good to tell 327 00:17:42,320 --> 00:17:44,557 the audience how do you see this 328 00:17:44,640 --> 00:17:49,437 diffusion occur in the utilization at the corporate level or at a government 329 00:17:49,520 --> 00:17:53,357 level that which ultimately then creates that demand which eliminates any fears 330 00:17:53,440 --> 00:17:55,837 and bubbles. >> Yeah. No, I think it's probably one of 331 00:17:55,920 --> 00:17:58,157 the the the big challenges with all of 332 00:17:58,240 --> 00:18:02,717 these new technologies is when uh work 333 00:18:02,800 --> 00:18:06,637 work artifact and work flow changes. 334 00:18:06,720 --> 00:18:10,637 uh that means we as firms have to change 335 00:18:10,720 --> 00:18:15,117 how we work. In fact, [clears throat] I remember meeting um uh the CEO of 336 00:18:15,200 --> 00:18:17,277 General Ali, you know, a few years back 337 00:18:17,360 --> 00:18:22,947 and he was describing he had joined um uh the firm, you know, prec 338 00:18:23,030 --> 00:18:25,997 >> [clears throat] >> era and uh uh and he was describing how, 339 00:18:26,080 --> 00:18:27,677 for example, they worked with their 340 00:18:27,760 --> 00:18:30,957 agents in the field uh with faxes, inter 341 00:18:31,040 --> 00:18:34,317 office memos and um and and suddenly the 342 00:18:34,400 --> 00:18:38,397 PC showed up and people would then put a spreadsheet in an email and send it 343 00:18:38,480 --> 00:18:40,717 around and the entire workflow and the 344 00:18:40,800 --> 00:18:43,197 work process has changed right so 345 00:18:43,280 --> 00:18:46,077 similarly I think with AI uh you are 346 00:18:46,160 --> 00:18:51,037 going to start seeing uh actual change in how workflow happens right I mean 347 00:18:51,120 --> 00:18:54,317 even [clears throat] in fact for me coming to Davos you know 348 00:18:54,400 --> 00:18:57,037 whatever 50 bilateral meetings I have 349 00:18:57,120 --> 00:19:00,077 preparing for those uh had a particular 350 00:19:00,160 --> 00:19:04,957 workflow right which is uh there is to be my field team would prepare notes and 351 00:19:05,040 --> 00:19:09,037 that would come to my HQ and that would get further refined and nothing had 352 00:19:09,120 --> 00:19:11,597 really changed right since I joined in 353 00:19:11,680 --> 00:19:16,637 '92 to essentially even a few years back. Whereas now I just go to co-pilot 354 00:19:16,720 --> 00:19:20,957 and say hey I'm meeting Larry. Please give me a brief and it comes back and 355 00:19:21,040 --> 00:19:25,037 gives me by the way the one nice thing is it gives me a 360 right? It knows 356 00:19:25,120 --> 00:19:29,437 what we're doing with you as a client what we're doing as a client of yours 357 00:19:29,520 --> 00:19:31,677 and everything in between [clears throat] as an investment. It's 358 00:19:31,760 --> 00:19:34,957 so it captures even information unlike 359 00:19:35,040 --> 00:19:39,197 anything else. In fact, what I do is I take that and immediately share that 360 00:19:39,280 --> 00:19:42,877 back with all my colleagues across all the functions. Right? Think about it. 361 00:19:42,960 --> 00:19:45,677 It's a complete inversion of how 362 00:19:45,760 --> 00:19:50,077 information is flowing in the organization. It's not like this classic 363 00:19:50,160 --> 00:19:53,357 we have an organization, we have departments, we have these 364 00:19:53,440 --> 00:19:55,437 specializations and the information 365 00:19:55,520 --> 00:19:58,557 trickles up. No, no, no. It actually it 366 00:19:58,640 --> 00:20:02,237 flattens the entire information flow. So 367 00:20:02,320 --> 00:20:04,877 once you start having that you have to 368 00:20:04,960 --> 00:20:08,317 redesign structurally. Uh so the current 369 00:20:08,400 --> 00:20:12,237 structure may not make sense. Um because 370 00:20:12,320 --> 00:20:16,077 you want people to be able to work in a 371 00:20:16,160 --> 00:20:18,637 way that allows them to have this 372 00:20:18,720 --> 00:20:23,597 information flow freely. So what all this leads me to if I had to sort of say 373 00:20:23,680 --> 00:20:25,597 what's the formula? The formula I think 374 00:20:25,680 --> 00:20:27,677 it starts with the mindset. So the 375 00:20:27,760 --> 00:20:31,917 mindset we as leaders should have is we 376 00:20:32,000 --> 00:20:34,957 need to think about changing the work 377 00:20:35,040 --> 00:20:39,517 the workflow with the technology then that needs skill set. [clears throat] So 378 00:20:39,600 --> 00:20:43,677 you can't sort of talk about this in the abstract. You got to use it like so if 379 00:20:43,760 --> 00:20:45,837 I'm not using the >> you have to trust it. 380 00:20:45,920 --> 00:20:49,997 >> You have to trust it. You have to use it. You have to learn even how to put 381 00:20:50,080 --> 00:20:53,997 the guard rails to trust it. Right? So you can't again you can't just be afraid 382 00:20:54,080 --> 00:20:56,877 of it. uh it's going to it's going to be 383 00:20:56,960 --> 00:20:59,677 diffused. So the question is as a firm 384 00:20:59,760 --> 00:21:02,877 you have to use it to learn how to even 385 00:21:02,960 --> 00:21:05,037 uh put the guard rails that allow you to 386 00:21:05,120 --> 00:21:08,237 be able to trust it. So skills uh so 387 00:21:08,320 --> 00:21:10,557 mindset skills the other big 388 00:21:10,640 --> 00:21:14,077 consideration um really is how do you 389 00:21:14,160 --> 00:21:17,197 make sure you have the data set that 390 00:21:17,280 --> 00:21:20,317 you're feeding like context like it's 391 00:21:20,400 --> 00:21:22,877 kind of like you have a new intelligence 392 00:21:22,960 --> 00:21:25,437 layer but the intelligence layer is only 393 00:21:25,520 --> 00:21:29,117 as good as the context you give it. So 394 00:21:29,200 --> 00:21:33,917 people describe it even as context engineering but that is what firms do 395 00:21:34,000 --> 00:21:37,997 right if you think about what do firms do it's all about the tacit knowledge we 396 00:21:38,080 --> 00:21:43,357 have by working as people in various department and moving paper and 397 00:21:43,440 --> 00:21:46,877 information so the question is how do you really have this AI also have that 398 00:21:46,960 --> 00:21:49,357 context so these are sort of some of the 399 00:21:49,440 --> 00:21:52,157 new things that have to percolate 400 00:21:52,240 --> 00:21:56,477 throughout an organization uh to take advantage in fact that's why I think you 401 00:21:56,560 --> 00:21:59,117 you're going to see that challenge of 402 00:21:59,200 --> 00:22:03,677 why am I not seeing immediate results in productivity because you have to do the 403 00:22:03,760 --> 00:22:06,877 hard work. In fact, that's why it's not going to be at some you know it's going 404 00:22:06,960 --> 00:22:10,077 to there are going to be firmwide differences. They're going to there 405 00:22:10,160 --> 00:22:13,037 could [clears throat] be sectorwide differences but it's going to 406 00:22:13,120 --> 00:22:17,997 fundamentally be because of the leadership will in an organization. 407 00:22:18,080 --> 00:22:22,877 >> Do you see the applications being used across large companies and medium and 408 00:22:22,960 --> 00:22:26,477 small companies or is it still the domain of mostly the large companies at 409 00:22:26,560 --> 00:22:28,717 this moment? I I think that what you're 410 00:22:28,800 --> 00:22:32,157 seeing is it's easier the because if you 411 00:22:32,240 --> 00:22:36,237 have a green sort of um uh you know if 412 00:22:36,320 --> 00:22:39,037 you start fresh it's easier to adopt 413 00:22:39,120 --> 00:22:43,997 these tools and you construct your organization knowing that these tools 414 00:22:44,080 --> 00:22:45,917 exist. So >> is it a barbell then? 415 00:22:46,000 --> 00:22:50,157 >> It is a barbell. So small companies that are just starting use that platform 416 00:22:50,240 --> 00:22:53,467 >> 100%. And I think in fact I would say even for large organizations there's 417 00:22:53,550 --> 00:22:56,717 [clears throat] a fundamental challenge right because unless and until your rate 418 00:22:56,800 --> 00:22:59,597 of change keeps up with >> right 419 00:22:59,680 --> 00:23:02,077 >> uh with with what is possible uh you're 420 00:23:02,160 --> 00:23:04,557 you're going to get schooled by someone 421 00:23:04,640 --> 00:23:09,277 small being able to achieve scale because of these tools. So but I think 422 00:23:09,360 --> 00:23:12,557 scale like large organizations have an inherent strength. You have the 423 00:23:12,640 --> 00:23:14,397 relationships, you have the data, you 424 00:23:14,480 --> 00:23:17,997 have um you have knowhow, but the bottom 425 00:23:18,080 --> 00:23:20,557 line is if you don't translate that with 426 00:23:20,640 --> 00:23:25,437 a new production function, uh then you really will be stuck. And so therefore 427 00:23:25,520 --> 00:23:29,037 the change management challenge for large organizations is going to be 428 00:23:29,120 --> 00:23:33,517 bigger. The structural challenge for small organizations of how to overcome 429 00:23:33,600 --> 00:23:36,717 scale issues is going to be harder. So it's sort of the two sides. in an 430 00:23:36,800 --> 00:23:38,237 interesting way. It's going to be a very 431 00:23:38,320 --> 00:23:41,917 competitively intense world uh where 432 00:23:42,000 --> 00:23:44,077 neither side like whether you're a new 433 00:23:44,160 --> 00:23:46,557 entrant or an incumbent can't take it as 434 00:23:46,640 --> 00:23:50,557 like I I can just coast. What about 435 00:23:50,640 --> 00:23:54,317 country to country? Are you seeing big differences in how the applications are 436 00:23:54,400 --> 00:23:57,357 being used? Is it is AI still the domain 437 00:23:57,440 --> 00:24:02,157 of developed countries or is it becoming rapidly a domain of all countries? I I 438 00:24:02,240 --> 00:24:04,237 I'm seeing there are two things I would 439 00:24:04,320 --> 00:24:06,717 say Larry as I travel around the world 440 00:24:06,800 --> 00:24:11,197 the quality of um whether it's the 441 00:24:11,280 --> 00:24:13,677 knowhow the software developers the 442 00:24:13,760 --> 00:24:15,277 startups 443 00:24:15,360 --> 00:24:19,037 um or even large or large organizations 444 00:24:19,120 --> 00:24:22,557 it's not that different it's fascinating you can show up in Jakarta you can show 445 00:24:22,640 --> 00:24:24,797 up in Istanbul you can show up in Mexico 446 00:24:24,880 --> 00:24:27,277 City it's not that different than 447 00:24:27,360 --> 00:24:29,837 showing up even in say Seattle or San 448 00:24:29,920 --> 00:24:32,077 Francisco right it's not uh for the 449 00:24:32,160 --> 00:24:35,197 first time just because access to what's 450 00:24:35,280 --> 00:24:39,577 happening uh is there that said at 451 00:24:39,660 --> 00:24:43,517 [clears throat] scale the commitment to 452 00:24:43,600 --> 00:24:46,557 using this the risk capital being there 453 00:24:46,640 --> 00:24:49,837 the large companies pushing it hard I 454 00:24:49,920 --> 00:24:53,437 mean I you know again the US you know is 455 00:24:53,520 --> 00:24:56,157 in fact if I compare it uh take 456 00:24:56,240 --> 00:24:59,437 financial sector if financial sector's 457 00:24:59,520 --> 00:25:01,117 adoption of the cloud 458 00:25:01,200 --> 00:25:03,997 versus AI night and day, right? Because 459 00:25:04,080 --> 00:25:07,357 in an interesting way, it's much faster 460 00:25:07,440 --> 00:25:11,837 uh when it comes to AI versus it was with the cloud and cloud because for a 461 00:25:11,920 --> 00:25:15,197 variety of reasons, >> regulatory issues too until the 462 00:25:15,280 --> 00:25:20,877 regulators allowed banks to bring their data off out of campus that was a big 463 00:25:20,960 --> 00:25:23,677 issue. >> Yeah. So I would say I think wherever [clears throat] 464 00:25:23,760 --> 00:25:26,637 you know so in the west in particular in 465 00:25:26,720 --> 00:25:30,717 the US uh there is clearly a real I 466 00:25:30,800 --> 00:25:33,357 would say more of an energy around it in 467 00:25:33,440 --> 00:25:36,237 terms of going and using it. Uh but it's 468 00:25:36,320 --> 00:25:41,597 sort of a lot more uniformly spreading around the world than any technology at 469 00:25:41,680 --> 00:25:44,237 least I've seen. But are you are [clears throat] you you mentioned about 470 00:25:44,320 --> 00:25:47,517 the power the grid is that going to be 471 00:25:47,600 --> 00:25:50,557 one of the determinants of of the 472 00:25:50,640 --> 00:25:53,117 accessibility if you do not have cheap 473 00:25:53,200 --> 00:25:56,397 power it the demand is costly 474 00:25:56,480 --> 00:25:59,597 >> 100%. So if you sort of look at the 475 00:25:59,680 --> 00:26:01,997 tokens per dollar per watt right which I 476 00:26:02,080 --> 00:26:03,517 think in some [clears throat] sense I 477 00:26:03,600 --> 00:26:07,197 would claim that GDP growth in any place 478 00:26:07,280 --> 00:26:11,117 will be directly correlated like you if you sort of buy my entire argument that 479 00:26:11,200 --> 00:26:15,037 look we've got a new commodity its tokens >> right 480 00:26:15,120 --> 00:26:19,677 >> and the job of every economy uh and every firm in the economy is to 481 00:26:19,760 --> 00:26:21,757 translate these tokens into economic 482 00:26:21,840 --> 00:26:24,317 growth then if you have a cheaper 483 00:26:24,400 --> 00:26:29,597 commodity it's better uh And so that's sort of what why there's tokens per 484 00:26:29,680 --> 00:26:33,277 dollar per watt. And by the way, there are many many elements to this, right? 485 00:26:33,360 --> 00:26:37,437 Which is uh it's not just um the production side. That's why I think even 486 00:26:37,520 --> 00:26:40,237 having the grid is important. Um uh 487 00:26:40,320 --> 00:26:42,317 construction costs, right? So if you 488 00:26:42,400 --> 00:26:45,437 like if you think about the total TCO uh 489 00:26:45,520 --> 00:26:49,757 everything it's like the how are you a cheap producer of energy? Can you build 490 00:26:49,840 --> 00:26:52,317 the data centers? Uh then what's the 491 00:26:52,400 --> 00:26:55,037 cost curve uh of the silicon and the 492 00:26:55,120 --> 00:26:57,757 systems? uh the and by the way look at 493 00:26:57,840 --> 00:27:00,797 the token pricing right token pricing 494 00:27:00,880 --> 00:27:04,557 basically drops by you know a half uh 495 00:27:04,640 --> 00:27:07,997 every 3 months uh I mean this is a so 496 00:27:08,080 --> 00:27:10,797 how so that that's why I think you can 497 00:27:10,880 --> 00:27:14,157 sort of really plot how you use the 498 00:27:14,240 --> 00:27:18,717 tokens to create surplus knowing that you have a commodity that's whose prices 499 00:27:18,800 --> 00:27:21,917 are just going to monotonically come down in a pretty fast curve 500 00:27:22,000 --> 00:27:24,957 >> we're sitting in Europe uh and there is 501 00:27:25,040 --> 00:27:27,197 a real fear 502 00:27:27,280 --> 00:27:29,277 because of the co Europe does not have 503 00:27:29,360 --> 00:27:32,317 its own power. It has to import mostly 504 00:27:32,400 --> 00:27:37,837 of its power. Um do you have any messages for Europe 505 00:27:37,920 --> 00:27:40,557 related to this? >> Yeah, I mean I think so there are two 506 00:27:40,640 --> 00:27:42,397 sets of things right. one is you know 507 00:27:42,480 --> 00:27:44,877 here we are in in Switzerland and I look 508 00:27:44,960 --> 00:27:47,837 at uh the the pharma or the financial 509 00:27:47,920 --> 00:27:50,637 sector you know obviously Switz they're 510 00:27:50,720 --> 00:27:54,877 uh they they do do a big job in in in this country as in in Europe but they're 511 00:27:54,960 --> 00:27:57,757 also international brands and international operations. So one thing 512 00:27:57,840 --> 00:28:00,797 that uh whenever I think about Europe is 513 00:28:00,880 --> 00:28:04,957 the Europeans are producing products and services that actually are going 514 00:28:05,040 --> 00:28:07,437 everywhere in the world and so therefore 515 00:28:07,520 --> 00:28:12,637 uh European competitiveness is about the competitiveness of their output globally 516 00:28:12,720 --> 00:28:15,997 not just in Europe. I think sometimes when you come to Europe there's a lot of 517 00:28:16,080 --> 00:28:18,637 conversation about just Europe. Uh but 518 00:28:18,720 --> 00:28:22,557 European economy is thrives and has you 519 00:28:22,640 --> 00:28:26,237 know thrived in the last whatever 200 years 300 years. the miracle of the west 520 00:28:26,320 --> 00:28:30,157 is fundamentally because of what has happened in Europe uh is because they 521 00:28:30,240 --> 00:28:33,597 were able to produce things that the world needed and so I would say that's 522 00:28:33,680 --> 00:28:35,757 number one and in order to do that you 523 00:28:35,840 --> 00:28:37,917 again I go back to the human capital 524 00:28:38,000 --> 00:28:40,717 here is just fantastic and world class 525 00:28:40,800 --> 00:28:43,757 uh you have to absolutely invest in uh 526 00:28:43,840 --> 00:28:48,397 producing uh you know having the energy and the tokens here which again you're 527 00:28:48,480 --> 00:28:52,237 attracting like as I said we're investing and others are investing uh 528 00:28:52,320 --> 00:28:56,317 the the data centers here so the question is what's that next generation 529 00:28:56,400 --> 00:29:00,557 of output that comes from here right I always think about the German middlear 530 00:29:00,640 --> 00:29:05,117 whenever I go to a jeweler or a dentist in the United States I'm surrounded by 531 00:29:05,200 --> 00:29:07,277 German middle right totally 532 00:29:07,360 --> 00:29:09,757 >> um it's just unbelievable engineering 533 00:29:09,840 --> 00:29:13,117 provice of that country uh and now the 534 00:29:13,200 --> 00:29:14,957 question and by the way that's the point 535 00:29:15,040 --> 00:29:17,517 that they are producing industrial 536 00:29:17,600 --> 00:29:21,277 products which today are built in into 537 00:29:21,360 --> 00:29:24,157 it all the intelligence as well that 538 00:29:24,240 --> 00:29:26,477 data right so I know whenever we come to 539 00:29:26,560 --> 00:29:28,557 Europe everyone's like talking about 540 00:29:28,640 --> 00:29:31,277 sovereignty and data this data that 541 00:29:31,360 --> 00:29:34,157 guess what Europe actually should be 542 00:29:34,240 --> 00:29:37,997 much more concerned about access to 543 00:29:38,080 --> 00:29:40,317 their industrial companies their 544 00:29:40,400 --> 00:29:43,197 financial services companies of data 545 00:29:43,280 --> 00:29:48,237 from you US and the rest of the world uh as opposed to just thinking that somehow 546 00:29:48,320 --> 00:29:51,597 by protecting Europe you're going to be competitive you are only going to be 547 00:29:51,680 --> 00:29:56,877 competitive if the products coming out of Europe are globally competitive, 548 00:29:56,960 --> 00:30:00,637 right? >> Um and so that's I think what needs to 549 00:30:00,720 --> 00:30:02,717 change. uh you know Europe has led in 550 00:30:02,800 --> 00:30:05,677 privacy that's fantastic has led in many 551 00:30:05,760 --> 00:30:10,237 aspects of even safety around AI and what have you and that's a feature uh 552 00:30:10,320 --> 00:30:14,477 that's great but you also have to complement it with by building locally 553 00:30:14,560 --> 00:30:16,877 and then also thinking globally what's 554 00:30:16,960 --> 00:30:19,357 the contribution this continent will 555 00:30:19,440 --> 00:30:21,677 make uh to the rest of the world which 556 00:30:21,760 --> 00:30:24,077 it has historically been a leader 557 00:30:24,160 --> 00:30:29,197 >> a leader so do you think the whole idea around sovereignty of data is that being 558 00:30:29,280 --> 00:30:33,117 misunderstood I I think that that when people talk 559 00:30:33,200 --> 00:30:35,437 about sovereignty 560 00:30:35,520 --> 00:30:41,277 first of all it's very important clearly and in a week like this it's more 561 00:30:41,360 --> 00:30:42,317 important 562 00:30:42,400 --> 00:30:46,077 >> uh but that said uh 563 00:30:46,160 --> 00:30:49,597 >> it is you have to kind of think about 564 00:30:49,680 --> 00:30:54,157 how what is sovereignty mean like for example in the AI era the topic that's 565 00:30:54,240 --> 00:30:56,477 least talked about but I feel will be 566 00:30:56,560 --> 00:30:59,677 most talked about in in this uh this 567 00:30:59,760 --> 00:31:01,757 calendar year will be the sovereignty of 568 00:31:01,840 --> 00:31:05,917 a firm. Just imagine if your firm you're 569 00:31:06,000 --> 00:31:09,517 not able to embed the tacit knowledge of 570 00:31:09,600 --> 00:31:13,677 the firm in a set of weights in a model 571 00:31:13,760 --> 00:31:16,717 that you control. By definition, you 572 00:31:16,800 --> 00:31:18,717 have no sovereignty. That means you're 573 00:31:18,800 --> 00:31:21,837 leaking enterprise value to some model 574 00:31:21,920 --> 00:31:25,837 company somewhere. In fact, that it's sort of fascinating that nobody's 575 00:31:25,920 --> 00:31:28,557 talking about that, right? It's like everybody's talking about everything 576 00:31:28,640 --> 00:31:31,757 else that is sort of you know outside of 577 00:31:31,840 --> 00:31:34,637 that whereas the most important thing is 578 00:31:34,720 --> 00:31:38,957 it really doesn't matter if you in fact the data center where it runs is the 579 00:31:39,040 --> 00:31:42,477 least important thing quite honestly but like even there first of all the data 580 00:31:42,560 --> 00:31:46,957 centers all are all over just because speed of light is a real constraint uh 581 00:31:47,040 --> 00:31:51,197 and so therefore the data centers will be spread yes uh you will have digital 582 00:31:51,280 --> 00:31:54,797 you'll be able to encrypt everything you'll be able to have the keys with you 583 00:31:54,880 --> 00:31:56,957 all of these are much more technically 584 00:31:57,040 --> 00:31:59,837 solve problems. But the one problem that 585 00:31:59,920 --> 00:32:03,037 will only be solved is by you having 586 00:32:03,120 --> 00:32:05,357 much more sovereignty over the you know 587 00:32:05,440 --> 00:32:08,157 tacid knowledge uh and control over the 588 00:32:08,240 --> 00:32:10,237 models and it's not a one-way enterprise 589 00:32:10,320 --> 00:32:13,197 value transfer. Um and so to me I think 590 00:32:13,280 --> 00:32:15,757 sovereignty requires real thought on 591 00:32:15,840 --> 00:32:18,477 what is it um you know control of 592 00:32:18,560 --> 00:32:22,237 destiny means that your your ability to 593 00:32:22,320 --> 00:32:26,797 produce something that is unique is preserved. uh David Ricardo was not 594 00:32:26,880 --> 00:32:29,117 wrong. There's comparative advantage in 595 00:32:29,200 --> 00:32:31,277 countries. Uh there is comparative 596 00:32:31,360 --> 00:32:34,317 advantage in firms that needs to be 597 00:32:34,400 --> 00:32:38,557 preserved even in the AI era. That's what'll give you real sovereignty. 598 00:32:38,640 --> 00:32:42,077 >> One last question. I know we're running out of time. 599 00:32:42,160 --> 00:32:46,077 Um in 5 years or 10 years, is there going to be one dominant model that 600 00:32:46,160 --> 00:32:50,477 we're all going to be using or are and how is Microsoft preparing for this? So 601 00:32:50,560 --> 00:32:51,677 are you going to be are we going to be 602 00:32:51,760 --> 00:32:55,037 using one model for for enterprise, one 603 00:32:55,120 --> 00:32:57,917 model for other other traits? 604 00:32:58,000 --> 00:33:00,157 >> You know, even in the last whatever 3 605 00:33:00,240 --> 00:33:02,317 years, four years that we've been at it, 606 00:33:02,400 --> 00:33:06,557 um the the reality at this point is it's 607 00:33:06,640 --> 00:33:10,077 a multimodel world, right? I mean the in 608 00:33:10,160 --> 00:33:14,477 fact if you think about it the in both 609 00:33:14,560 --> 00:33:18,397 there are going to be multiple models 610 00:33:18,480 --> 00:33:21,597 and the trick is really how do you take 611 00:33:21,680 --> 00:33:24,557 advantage of these multiple models and 612 00:33:24,640 --> 00:33:27,037 in fact build your own model by 613 00:33:27,120 --> 00:33:29,837 distilling these right uh so think of 614 00:33:29,920 --> 00:33:33,117 these models uh that you orchestrate to 615 00:33:33,200 --> 00:33:37,997 build your own model and more importantly you do what is described as 616 00:33:38,080 --> 00:33:39,517 orchest orchestration or harness 617 00:33:39,600 --> 00:33:42,317 engineering. So the IP of any 618 00:33:42,400 --> 00:33:45,517 application or any firm is how do you 619 00:33:45,600 --> 00:33:50,317 use all these models with context engineering or your data. Yes. 620 00:33:50,400 --> 00:33:54,477 >> Right. So it's that three parts. So can I bring in all the models by the way uh 621 00:33:54,560 --> 00:33:57,037 which is closed source, open source, 622 00:33:57,120 --> 00:34:00,797 build my own model, orchestrate them and 623 00:34:00,880 --> 00:34:04,397 feed it my data to change the trajectory 624 00:34:04,480 --> 00:34:07,117 of some outcome that I care about. 625 00:34:07,200 --> 00:34:09,597 That's it. That's the entire picture. So 626 00:34:09,680 --> 00:34:14,237 you can do it in like oh I produce a particular product or service. Uh first 627 00:34:14,320 --> 00:34:18,797 I got to do better better better job in sales or better job in R&D or better job 628 00:34:18,880 --> 00:34:22,717 in finance or what have you. And you take that outcome and then you say can I 629 00:34:22,800 --> 00:34:25,197 use all the models orchestrate them and 630 00:34:25,280 --> 00:34:27,597 feed it my context and then in as a 631 00:34:27,680 --> 00:34:30,557 result of it the reasoning traces are 632 00:34:30,640 --> 00:34:33,437 really leading to some capability and 633 00:34:33,520 --> 00:34:36,637 models that I control as my IP. As long 634 00:34:36,720 --> 00:34:41,117 as firms can answer that question, they're going to be getting ahead. 635 00:34:41,200 --> 00:34:43,597 >> Ladies and gentlemen, let's uh thank 636 00:34:43,680 --> 00:34:50,360 Satya, my friend. Thank you [applause] for 637 00:34:51,200 --> 00:34:54,957 and hopefully this is the beginning of many great dialogue and conversations 638 00:34:55,040 --> 00:34:58,077 here at the World Academic Forum. Thank you everyone. 639 00:34:58,160 --> 00:35:01,320 >> Thank you.