WEBVTT 00:00:00.080 --> 00:00:02.077 is everything. And so the way it happens 00:00:02.160 --> 00:00:06.237 is let's sort of unpack this uh on the 00:00:06.320 --> 00:00:08.397 supply side what needs to happen in each 00:00:08.480 --> 00:00:11.437 country is the tokens per dollar per 00:00:11.520 --> 00:00:14.957 watt have to sort of monotonically get 00:00:15.040 --> 00:00:19.197 more efficient and better. Right? So to some degree even what we're trying to do 00:00:19.280 --> 00:00:23.837 with the investments the two firms are doing uh around the world is to just say 00:00:23.920 --> 00:00:25.517 that like let's make sure that the 00:00:25.600 --> 00:00:28.157 supply is there which is uh everything 00:00:28.240 --> 00:00:33.357 from the chips on down ultimately to these token factories that get deployed 00:00:33.440 --> 00:00:37.517 everywhere. By the way there's not one token factory. This token factory is the 00:00:37.600 --> 00:00:41.117 first thing that's going to be diffused all around the world. It's just like 00:00:41.200 --> 00:00:44.157 electricity right? you just need a 00:00:44.240 --> 00:00:48.877 ubiquitous grid of uh energy and tokens 00:00:48.960 --> 00:00:51.517 that then will power the rest of the 00:00:51.600 --> 00:00:53.437 economy right so that's I think one side 00:00:53.520 --> 00:00:56.877 of it then the demand side of this is a 00:00:56.960 --> 00:00:59.917 little bit like every firm has to start 00:01:00.000 --> 00:01:03.197 by using it if I look back even you know 00:01:03.280 --> 00:01:08.477 when the PCs first came out or the personal computing era started I I loved 00:01:08.560 --> 00:01:12.797 you know I think Jobs had a nice metaphor he called it the bicycle for the 00:01:12.880 --> 00:01:16.957 uh Bill had a metaphor which I remember was like information at your fingertips. 00:01:17.040 --> 00:01:20.637 Right? These two metaphors were great like which allowed us to say that's what 00:01:20.720 --> 00:01:22.797 it is. It's a tool that I will use to 00:01:22.880 --> 00:01:24.877 get information at my fingertips. I'll 00:01:24.960 --> 00:01:28.637 use it as a cognitive amplifier. Now I 00:01:28.720 --> 00:01:31.357 think that's what we have but you know 00:01:31.440 --> 00:01:33.917 10x 100x right? So in some sense you as 00:01:34.000 --> 00:01:37.677 a as every knowledge worker you now have 00:01:37.760 --> 00:01:41.837 access to infinite minds. That's the way I think about it. Right? So there's a 00:01:41.920 --> 00:01:44.317 Turing award winner uh called Raj Ready 00:01:44.400 --> 00:01:49.037 who had this nice metaphor of AI and he had this long before uh even generative 00:01:49.120 --> 00:01:53.197 AI he said either either it's a cognitive amplifier or it's a guardian 00:01:53.280 --> 00:01:55.997 angel right so if you think of AI as 00:01:56.080 --> 00:01:59.037 that um then that in the global 00:01:59.120 --> 00:02:03.117 workforce right when a doctor can get to 00:02:03.200 --> 00:02:07.117 a patient spend more time with the patient because the AI is doing the 00:02:07.200 --> 00:02:09.357 transcription and entering the records 00:02:09.440 --> 00:02:12.237 in the EMR system entering the right 00:02:12.320 --> 00:02:14.957 billing code so that the health care uh 00:02:15.040 --> 00:02:17.117 you know industry is better served 00:02:17.200 --> 00:02:20.157 across the payer the provider and uh and 00:02:20.240 --> 00:02:23.037 the patient ultimately right that's an 00:02:23.120 --> 00:02:25.117 an outcome that I think all of us can 00:02:25.200 --> 00:02:28.077 benefit from so I feel ultimately it's 00:02:28.160 --> 00:02:30.637 going to require real leadership on the 00:02:30.720 --> 00:02:33.197 private sector and the public sector to 00:02:33.280 --> 00:02:35.997 ensure the diffusion happens and the one 00:02:36.080 --> 00:02:39.117 thing I'll mention Larry is skilling 00:02:39.200 --> 00:02:43.117 right so In some sense the thing that 00:02:43.200 --> 00:02:46.877 diffusion is very strongly correlated to 00:02:46.960 --> 00:02:49.837 one thing alone which is how broadly are 00:02:49.920 --> 00:02:54.077 people skilled in using this. Um and 00:02:54.160 --> 00:02:56.317 interestingly enough, I think if mobile 00:02:56.400 --> 00:02:59.677 has taught us one thing is it it's 00:02:59.760 --> 00:03:03.597 actually distinct from what happened in the PC, right? Uh I remember even 00:03:03.680 --> 00:03:05.997 growing up in the global south, uh there 00:03:06.080 --> 00:03:08.797 used to be a real relationship between 00:03:08.880 --> 00:03:12.157 learning Excel skills or word skills and 00:03:12.240 --> 00:03:16.397 getting a job. Um you know, right now um 00:03:16.480 --> 00:03:20.557 what's the model in in mobile? It's kind of created the same opportunity, but 00:03:20.640 --> 00:03:22.717 it's been a lot more consumptionled. 00:03:22.800 --> 00:03:27.037 It's these creator economy and what have you, but it has not been about sort of, 00:03:27.120 --> 00:03:30.717 oh wow, here's how you get a healthcare job or here is how you get a finance job 00:03:30.800 --> 00:03:32.877 or here is how you get, you know, you 00:03:32.960 --> 00:03:35.517 get ahead professionally. 00:03:35.600 --> 00:03:38.637 Um, and that needs to come back, right? 00:03:38.720 --> 00:03:40.477 People need to say, I pick up this AI 00:03:40.560 --> 00:03:43.437 skill and now I'm a better provider of 00:03:43.520 --> 00:03:47.857 some product or service in the real economy. So it's it's very easy 00:03:47.940 --> 00:03:51.517 [clears throat] to see how mobile and the diffusion of mobile how it 00:03:51.600 --> 00:03:53.757 transformed economies especially in the 00:03:53.840 --> 00:03:57.117 global south. How does how do this you 00:03:57.200 --> 00:03:59.277 know to me I I I just read a research 00:03:59.360 --> 00:04:04.237 report that said the applications for AI so far are 00:04:04.320 --> 00:04:08.877 heavily weighted towards those who are educated or educated economies. 00:04:08.960 --> 00:04:12.957 And so does that create that you know more of a bifurcation a more 00:04:13.040 --> 00:04:15.277 polarization? How do we ensure that that 00:04:15.360 --> 00:04:18.317 that diffusion is spread evenly? How do 00:04:18.400 --> 00:04:19.957 we make sure that we're not 00:04:20.040 --> 00:04:25.437 [clears throat] leaving major portions of society or the world behind because I 00:04:25.520 --> 00:04:27.837 think that's that's going to [clears throat] be the big issue for us 00:04:27.920 --> 00:04:30.077 going forward. >> Yeah. So, it's it's it's interesting, 00:04:30.160 --> 00:04:33.837 right? This is one of those times when 00:04:33.920 --> 00:04:37.357 uh by definition and because of the 00:04:37.440 --> 00:04:41.357 rails that have been established you know as you said right which is 00:04:41.440 --> 00:04:44.237 >> uh what's happened with mobile as well 00:04:44.320 --> 00:04:48.717 as what's happened with um uh you know essentially connectivity 00:04:48.800 --> 00:04:52.317 >> right >> you have the ability to sort of deliver 00:04:52.400 --> 00:04:55.677 the tokens pretty evenly around the world >> right 00:04:55.760 --> 00:05:00.637 >> a lot more so than let's say uh the PC era or even the beginning of the mobile 00:05:00.720 --> 00:05:04.637 era, right? Because it took a long time for even the mo smartphone in particular 00:05:04.720 --> 00:05:09.277 to penetrate um all of the world. Whereas now it's 00:05:09.360 --> 00:05:11.517 not the case, right? These models and 00:05:11.600 --> 00:05:14.637 their outputs are pretty much available 00:05:14.720 --> 00:05:18.797 everywhere. And so the question to me is 00:05:18.880 --> 00:05:22.557 what's the use cases that make sense, right? It's one, in fact, one of the 00:05:22.640 --> 00:05:24.637 demos I always go back to. I think this 00:05:24.720 --> 00:05:27.197 was even in the beginning of 23 was a 00:05:27.280 --> 00:05:30.077 rural Indian farmer was able to use a 00:05:30.160 --> 00:05:33.357 bot built on I think a very early GPT3 00:05:33.440 --> 00:05:37.357 or 25 even uh essentially to reason over 00:05:37.440 --> 00:05:39.437 some farm subsidies that he had heard 00:05:39.520 --> 00:05:41.997 about in a local language and had it 00:05:42.080 --> 00:05:45.597 even in that very early days uh have it 00:05:45.680 --> 00:05:47.597 even show some agentic behavior, right? 00:05:47.680 --> 00:05:49.517 Like go complete a form for me. So in 00:05:49.600 --> 00:05:51.837 some sense it took you know it brought 00:05:51.920 --> 00:05:56.637 back agency to someone uh who perhaps 00:05:56.720 --> 00:06:01.197 didn't have that because the technology was so much more accessible. So I I do 00:06:01.280 --> 00:06:03.357 think it's in our hands even in the 00:06:03.440 --> 00:06:07.437 global south to use it uh to create I 00:06:07.520 --> 00:06:10.077 would say more of that opportunity where 00:06:10.160 --> 00:06:13.837 there isn't one. Um but I think what the 00:06:13.920 --> 00:06:17.677 necessary [clears throat] conditions still are do you have uh the 00:06:17.760 --> 00:06:22.317 capital investment being put in uh do you even have an environment for capital 00:06:22.400 --> 00:06:26.317 because in an interesting way we are for example as hyperscalers investing all 00:06:26.400 --> 00:06:30.557 over right including the global south uh so as long as there's an environment 00:06:30.640 --> 00:06:33.517 which attracts the capital investment >> and you see the demand 00:06:33.600 --> 00:06:36.397 >> and you see and then yeah the demand is there 00:06:36.480 --> 00:06:40.717 >> um and so the question is how do you have a set of policies that allow for 00:06:40.800 --> 00:06:45.277 both the capital to come in for it to find nexus with there are certain things 00:06:45.360 --> 00:06:49.197 by the way private capital can do certain things that public capital only 00:06:49.280 --> 00:06:52.237 can do for example the grid right it's 00:06:52.320 --> 00:06:57.197 not I mean grid in most countries is sort of fundamentally driven by governments 00:06:57.280 --> 00:07:00.397 >> and public >> and public and so if you don't have a 00:07:00.480 --> 00:07:04.397 sophisticated sort or rather if you don't have a a real approach to 00:07:04.480 --> 00:07:08.797 modernizing the grid uh that will hold things back I mean there's a lot of talk 00:07:08.880 --> 00:07:12.237 about behind the meter and so on and Yes, there's some amount of that we can 00:07:12.320 --> 00:07:14.877 do ourselves. >> We can do that in the US. Many countries can't. 00:07:14.960 --> 00:07:16.477 >> Exactly. And it's not long-term 00:07:16.560 --> 00:07:18.397 scalable, right? I mean, like to me, a 00:07:18.480 --> 00:07:21.917 long-term scalable solution is to have 00:07:22.000 --> 00:07:26.717 uh you know, all of these token factories, part of the real economy, 00:07:26.800 --> 00:07:31.357 connected to the grid, connected to the telco network, delivering uh just like 00:07:31.440 --> 00:07:33.357 we delivered bits, you have to deliver 00:07:33.440 --> 00:07:36.397 tokens plus bits. Um and that's kind of 00:07:36.480 --> 00:07:40.717 what's going to drive an at scale whether it's in the global south or in 00:07:40.800 --> 00:07:44.077 uh on the developed world. >> So so many people talk about there may 00:07:44.160 --> 00:07:46.077 be an AI bubble. I mean the most 00:07:46.160 --> 00:07:47.437 important thing that we see as an 00:07:47.520 --> 00:07:51.597 investor is the the democratization of 00:07:51.680 --> 00:07:54.477 technology is and the diffusion of that 00:07:54.560 --> 00:07:57.037 technology really does then transform 00:07:57.120 --> 00:08:02.157 the demand and the the companies or the countries that diffuse it fastest are 00:08:02.240 --> 00:08:03.677 going to be the ultimate winners. not 00:08:03.760 --> 00:08:05.517 the technology creator. 00:08:05.600 --> 00:08:09.757 >> That's that's you know it's it's it's 00:08:09.840 --> 00:08:12.157 for this not to be a bubble 00:08:12.240 --> 00:08:12.717 >> y 00:08:12.800 --> 00:08:16.957 >> by definition it requires that the 00:08:17.040 --> 00:08:19.197 benefits of this are much more evenly 00:08:19.280 --> 00:08:22.877 spread. I mean I think a telltel sign of 00:08:22.960 --> 00:08:26.077 if it's a bubble would be if all we're 00:08:26.160 --> 00:08:29.597 talking about are the tech firms. Uh 00:08:29.680 --> 00:08:34.397 right? If uh all we talk about is what's happening to the technology side that 00:08:34.480 --> 00:08:38.157 then that's by day you know it's just purely supply side >> right 00:08:38.240 --> 00:08:42.557 >> uh ultimately if we are not talking about wow here is a drug comp you know 00:08:42.640 --> 00:08:46.237 drug that was sort of uh brought into the market that's super successful 00:08:46.320 --> 00:08:50.797 because it was uh AI accelerated the clinical trial it's not even the magical 00:08:50.880 --> 00:08:52.797 molecule right it's kind of even the 00:08:52.880 --> 00:08:55.837 rest of what is uh needed in order to 00:08:55.920 --> 00:08:58.557 make something much more relevant right 00:08:58.640 --> 00:09:03.757 um and So the more [clears throat] we have uh and by the way it's happening 00:09:03.840 --> 00:09:08.317 right. So I'm not sort of saying that that's why I'm much more confident that 00:09:08.400 --> 00:09:10.717 this is a technology that will in fact 00:09:10.800 --> 00:09:14.157 build on the rails of cloud and mobile 00:09:14.240 --> 00:09:17.327 diffuse faster and bend the productivity 00:09:17.410 --> 00:09:21.437 [clears throat] curve and bring local surplus and 00:09:21.520 --> 00:09:23.997 economic growth all around the world. 00:09:24.080 --> 00:09:28.317 Not just economic growth driven by capital expenses. 00:09:28.400 --> 00:09:33.277 uh right because that's it's a narrow point in time uh calculation is 00:09:33.360 --> 00:09:35.837 >> right now that's what we're seeing more >> that's what we're seeing you know you 00:09:35.920 --> 00:09:37.357 know in in the developed world in 00:09:37.440 --> 00:09:40.957 particular uh but remember my capital 00:09:41.040 --> 00:09:44.477 like that the one thing that you know is definitely we spending a lot of it in 00:09:44.560 --> 00:09:47.757 the United States but 50% of it is also all over the world 00:09:47.840 --> 00:09:51.677 >> right >> um and so interesting enough it depends 00:09:51.760 --> 00:09:56.077 on uh demand all over the world and the demand all over the world will only be 00:09:56.160 --> 00:09:59.597 there if there is local surplus plus all over the world. And so that's sort of 00:09:59.680 --> 00:10:04.317 the way I see the equation. >> So let's drill down a little more. As AI 00:10:04.400 --> 00:10:08.397 diffuses, obviously organizations, companies, 00:10:08.480 --> 00:10:10.797 governments are going to have to evolve. 00:10:10.880 --> 00:10:14.077 Now getting to the demand side. So how 00:10:14.160 --> 00:10:15.517 do you think the structure of 00:10:15.600 --> 00:10:19.437 organizations is changing in an AI world 00:10:19.520 --> 00:10:22.797 across roles, across teams, management? 00:10:22.880 --> 00:10:26.397 I'm I'm sure um Microsoft has evolved 00:10:26.480 --> 00:10:29.277 itself. So it probably be good to tell 00:10:29.360 --> 00:10:31.597 the audience how do you see this 00:10:31.680 --> 00:10:36.477 diffusion occur in the utilization at the corporate level or at a government 00:10:36.560 --> 00:10:40.237 level that which ultimately then creates that demand which eliminates any fears 00:10:40.320 --> 00:10:42.797 it bubbles. >> Yeah. No, I think it it's probably one 00:10:42.880 --> 00:10:45.117 of the the the big challenges with all 00:10:45.200 --> 00:10:48.637 of these new technologies is when uh 00:10:48.720 --> 00:10:52.157 work work artifact and work flow 00:10:52.240 --> 00:10:56.957 changes. Uh the means we as firms have 00:10:57.040 --> 00:10:59.277 to change how we work. In fact, I 00:10:59.360 --> 00:11:02.157 remember meeting um uh the CEO of 00:11:02.240 --> 00:11:04.317 General Ali, you know, a few years back 00:11:04.400 --> 00:11:09.987 and he was describing he had joined um uh the firm, you know, prec 00:11:10.070 --> 00:11:13.037 >> [clears throat] >> era and uh uh and he was describing how, 00:11:13.120 --> 00:11:14.717 for example, they worked with their 00:11:14.800 --> 00:11:17.997 agents in the field uh with faxes, inter 00:11:18.080 --> 00:11:21.277 office memos and um and and suddenly the 00:11:21.360 --> 00:11:25.437 PC showed up and people would then put a spreadsheet in an email and send it 00:11:25.520 --> 00:11:27.757 around and the entire workflow and the 00:11:27.840 --> 00:11:30.237 work process has changed right so 00:11:30.320 --> 00:11:33.117 similarly I think with AI uh you are 00:11:33.200 --> 00:11:38.077 going to start seeing uh actual change in how workflow happens right I mean 00:11:38.160 --> 00:11:41.277 even [clears throat] in fact for me coming to Davos you know 00:11:41.360 --> 00:11:44.077 whatever 50 bilateral meetings I have 00:11:44.160 --> 00:11:47.117 preparing for those uh had a particular 00:11:47.200 --> 00:11:51.997 workflow right which is uh there is to be my field team would prepare notes and 00:11:52.080 --> 00:11:56.077 that would come to my HQ and that would get further refined and nothing had 00:11:56.160 --> 00:11:58.637 really changed right since I joined in 00:11:58.720 --> 00:12:03.677 '92 to essentially even a few years back. Whereas now I just go to co-pilot 00:12:03.760 --> 00:12:07.997 and say hey I'm meeting Larry please give me a brief and it comes back and 00:12:08.080 --> 00:12:12.237 gives me by the way the one nice thing is it gives me a 360 right it knows what 00:12:12.320 --> 00:12:16.797 we're doing with you as a client what we're doing as a client of yours and 00:12:16.880 --> 00:12:20.797 everything in between [clears throat] as an investment it's so it captures even 00:12:20.880 --> 00:12:23.037 information unlike anything else in fact 00:12:23.120 --> 00:12:27.917 what I do is I take that and immediately share that back with all my colleagues 00:12:28.000 --> 00:12:32.317 across all the functions right think about it it's a complete inversion of 00:12:32.400 --> 00:12:37.117 how information is flowing in the organization. It's not like this classic 00:12:37.200 --> 00:12:40.397 we have an organization, we have departments, we have these 00:12:40.480 --> 00:12:42.477 specializations and the information 00:12:42.560 --> 00:12:45.597 trickles up. No, no, no. It actually it 00:12:45.680 --> 00:12:49.277 flattens the entire information flow. So 00:12:49.360 --> 00:12:51.917 once you start having that you have to 00:12:52.000 --> 00:12:55.357 redesign structurally. Uh so the current 00:12:55.440 --> 00:12:59.277 structure may not make sense. um because 00:12:59.360 --> 00:13:03.117 you want people to be able to work in a 00:13:03.200 --> 00:13:05.597 way that allows them to have this 00:13:05.680 --> 00:13:08.557 information flow freely. So what all 00:13:08.640 --> 00:13:12.637 this leads me to if I had to sort of say what's the formula the formula I think 00:13:12.720 --> 00:13:14.717 it starts with the mindset. So the 00:13:14.800 --> 00:13:18.957 mindset we as leaders should have is we 00:13:19.040 --> 00:13:21.997 need to think about changing the work 00:13:22.080 --> 00:13:25.117 the workflow with the technology then 00:13:25.200 --> 00:13:29.357 that needs skill set. So you can't sort of talk about this in the abstract. You 00:13:29.440 --> 00:13:32.877 got to use it. Like so if I'm not using the >> you have to trust it. 00:13:32.960 --> 00:13:37.037 >> You have to trust it. You have to use it. You have to learn even how to put 00:13:37.120 --> 00:13:41.037 the guard rails to trust it. Right? So you can't again you can't just be afraid 00:13:41.120 --> 00:13:43.917 of it. Uh it's going to it's going to be 00:13:44.000 --> 00:13:46.717 diffused. So the question is as a firm 00:13:46.800 --> 00:13:49.917 you have to use it to learn how to even 00:13:50.000 --> 00:13:51.997 uh put the guard rails that allow you to 00:13:52.080 --> 00:13:55.277 be able to trust it. So skills uh so 00:13:55.360 --> 00:13:57.597 mindset skills the other big 00:13:57.680 --> 00:14:01.117 consideration u really is how do you 00:14:01.200 --> 00:14:04.237 make sure you have the data set that 00:14:04.320 --> 00:14:07.357 you're feeding like context like it's 00:14:07.440 --> 00:14:09.917 kind of like you have a new intelligence 00:14:10.000 --> 00:14:12.477 layer but the intelligence layer is only 00:14:12.560 --> 00:14:16.157 as good as the context you give it. So 00:14:16.240 --> 00:14:20.957 people describe it even as context engineering but that is what firms do 00:14:21.040 --> 00:14:24.957 right if you think about what do firms do it's all about the tacet knowledge we 00:14:25.040 --> 00:14:27.917 have by working as people in various 00:14:28.000 --> 00:14:31.997 departments and moving paper and information so the question is how do 00:14:32.080 --> 00:14:33.917 you really have this AI also have that 00:14:34.000 --> 00:14:36.397 context so these are sort of some of the 00:14:36.480 --> 00:14:39.197 new things that have to percolate 00:14:39.280 --> 00:14:43.517 throughout an organization uh to take advantage in fact that's why I think you 00:14:43.600 --> 00:14:46.157 you're going to see that challenge of 00:14:46.240 --> 00:14:50.717 why am I not seeing immediate results in productivity because you have to do the 00:14:50.800 --> 00:14:53.917 hard work. In fact, that's why it's not going to be at some you know it's going 00:14:54.000 --> 00:14:57.117 to there going to be firmwide differences. They're going to there 00:14:57.200 --> 00:15:00.077 could be [clears throat] sectorwide differences but it's going to 00:15:00.160 --> 00:15:05.037 fundamentally be because of the leadership will in an organization. 00:15:05.120 --> 00:15:09.917 >> Do you see the applications being used across large companies and medium and 00:15:10.000 --> 00:15:13.517 small companies or is it still the domain of mostly the large companies at 00:15:13.600 --> 00:15:15.757 this moment? I I think that what you're 00:15:15.840 --> 00:15:19.197 seeing is it's easier the because if you 00:15:19.280 --> 00:15:23.277 have a green sort of um uh you know if 00:15:23.360 --> 00:15:26.077 you start fresh it's easier to adopt 00:15:26.160 --> 00:15:30.957 these tools and you construct your organization knowing that these tools 00:15:31.040 --> 00:15:32.957 exist. So >> is it a barbell then? 00:15:33.040 --> 00:15:37.197 >> It is a barbell. So small companies that are just starting use that platform 00:15:37.280 --> 00:15:40.507 >> 100%. And I think in fact I would say even for large organization there's 00:15:40.590 --> 00:15:43.757 [clears throat] a fundamental challenge right because unless and until your rate 00:15:43.840 --> 00:15:46.637 of change keeps up with >> right 00:15:46.720 --> 00:15:49.117 >> uh with with what is possible uh you 00:15:49.200 --> 00:15:51.597 you're going to get schooled by someone 00:15:51.680 --> 00:15:56.317 small being able to achieve scale because of these tools. So but I think 00:15:56.400 --> 00:15:59.597 scale large organizations have an inherent strength. You have the 00:15:59.680 --> 00:16:01.437 relationships, you have the data, you 00:16:01.520 --> 00:16:04.957 have um you have knowhow, but the bottom 00:16:05.040 --> 00:16:07.597 line is if you don't translate that with 00:16:07.680 --> 00:16:12.477 a new production function, uh then you really will be stuck. And so therefore 00:16:12.560 --> 00:16:15.997 the change management challenge for large organizations is going to be 00:16:16.080 --> 00:16:20.557 bigger. The structural challenge for small organizations of how to overcome 00:16:20.640 --> 00:16:23.757 scale issues is going to be harder. So it's sort of the two sides in an 00:16:23.840 --> 00:16:25.277 interesting way. It's going to be a very 00:16:25.360 --> 00:16:28.957 competitively intense world uh where 00:16:29.040 --> 00:16:31.117 neither side like whether you're a new 00:16:31.200 --> 00:16:33.597 entrant or an incumbent can't take it as 00:16:33.680 --> 00:16:37.597 like I I can just coast. What about 00:16:37.680 --> 00:16:41.357 country to country? Are you seeing big differences in how the applications are 00:16:41.440 --> 00:16:44.397 being used? Is it is AI still the domain 00:16:44.480 --> 00:16:49.197 of developed countries or is it becoming rapidly a domain of all countries? I I 00:16:49.280 --> 00:16:50.877 I'm seeing there are [clears throat] two 00:16:50.960 --> 00:16:52.637 things I would say Larry as I travel 00:16:52.720 --> 00:16:56.957 around the world the quality of um 00:16:57.040 --> 00:16:59.437 whether it's the knowhow the software 00:16:59.520 --> 00:17:02.317 developers the startups 00:17:02.400 --> 00:17:06.077 um or even large or large organizations 00:17:06.160 --> 00:17:09.597 it's not that different it's fascinating you can show up in Jakarta you can show 00:17:09.680 --> 00:17:11.757 up in Istanbul you can show up in Mexico 00:17:11.840 --> 00:17:14.317 City it's not that different than 00:17:14.400 --> 00:17:16.877 showing up even in say Seattle or San 00:17:16.960 --> 00:17:19.117 Francisco right it's not uh for the 00:17:19.200 --> 00:17:22.157 first time just because access to what's 00:17:22.240 --> 00:17:26.617 happening uh is there that said at 00:17:26.700 --> 00:17:30.557 [clears throat] scale the commitment to 00:17:30.640 --> 00:17:33.597 using this the risk capital being there 00:17:33.680 --> 00:17:36.877 the large companies pushing it hard I 00:17:36.960 --> 00:17:40.397 mean I you know again the US you know is 00:17:40.480 --> 00:17:43.117 in fact if I compare it uh take 00:17:43.200 --> 00:17:46.397 financial sector if financial sector's 00:17:46.480 --> 00:17:48.157 adoption of the cloud 00:17:48.240 --> 00:17:51.037 versus AI night and day, right? Because 00:17:51.120 --> 00:17:54.397 in an interesting way, it's much faster 00:17:54.480 --> 00:17:58.877 uh when it comes to AI versus it was with the cloud and cloud because for a 00:17:58.960 --> 00:18:02.237 variety of reasons >> and regulatory issues too until the 00:18:02.320 --> 00:18:07.837 regulators allowed banks to bring their data off out of campus that was a big 00:18:07.920 --> 00:18:10.717 issue. >> Yeah. So I would say I think wherever [clears throat] 00:18:10.800 --> 00:18:13.677 you know so in the west in particular in 00:18:13.760 --> 00:18:17.677 the US uh there is clearly a real I 00:18:17.760 --> 00:18:20.397 would say more of an energy around it in 00:18:20.480 --> 00:18:23.277 terms of going and using it uh but it's 00:18:23.360 --> 00:18:28.557 sort of a lot more uniformly spreading around the world than any technology at 00:18:28.640 --> 00:18:30.637 least I've seen >> but are you [clears throat] you 00:18:30.720 --> 00:18:33.997 mentioned about the power the grid is 00:18:34.080 --> 00:18:35.597 that going to be one of the determinants 00:18:35.680 --> 00:18:39.197 of of the accessibility If you do not 00:18:39.280 --> 00:18:43.437 have cheap power it the demand is costly 00:18:43.520 --> 00:18:46.637 >> 100%. So if you sort of look at the 00:18:46.720 --> 00:18:49.037 tokens per dollar per watt right which I 00:18:49.120 --> 00:18:50.557 think in some [clears throat] sense I 00:18:50.640 --> 00:18:54.157 would claim that GDP growth in any place 00:18:54.240 --> 00:18:58.157 will be directly correlated like you if you sort of buy my entire argument that 00:18:58.240 --> 00:19:01.997 look we've got a new commodity its tokens >> right 00:19:02.080 --> 00:19:06.717 >> and the job of every economy uh and every firm in the economy is to 00:19:06.800 --> 00:19:08.717 translate these tokens into economic 00:19:08.800 --> 00:19:11.277 growth then if you have a cheaper 00:19:11.360 --> 00:19:16.557 commodity it's better uh And so that's sort of what why there's tokens per 00:19:16.640 --> 00:19:20.317 dollar per watt. And by the way, there are many many elements to this, right? 00:19:20.400 --> 00:19:24.477 Which is uh it's not just um the production side. That's why I think even 00:19:24.560 --> 00:19:27.277 having the grid is important. Um uh 00:19:27.360 --> 00:19:29.277 construction costs, right? So if you 00:19:29.360 --> 00:19:32.237 like if you think about the total TCO 00:19:32.320 --> 00:19:36.557 uh everything it's like the how are you a cheap producer of energy? Can you 00:19:36.640 --> 00:19:39.037 build the data centers? Uh then what's 00:19:39.120 --> 00:19:41.997 the cost curve uh of the silicon and the 00:19:42.080 --> 00:19:44.797 systems? uh the and by the way look at 00:19:44.880 --> 00:19:47.757 the token pricing right token pricing 00:19:47.840 --> 00:19:51.597 basically drops by you know a half uh 00:19:51.680 --> 00:19:54.717 every three months uh I mean this is a 00:19:54.800 --> 00:19:57.357 so how so that that's why I think you 00:19:57.440 --> 00:20:01.197 can sort of really plot how you use the 00:20:01.280 --> 00:20:05.757 tokens to create surplus knowing that you have a commodity that's whose prices 00:20:05.840 --> 00:20:08.877 are just going to monotonically come down in a pretty fast curve 00:20:08.960 --> 00:20:11.997 >> we're sitting in Europe uh and there is 00:20:12.080 --> 00:20:14.237 a real fear 00:20:14.320 --> 00:20:16.557 because of the Europe does not have its 00:20:16.640 --> 00:20:19.437 own power. It has to import mostly of 00:20:19.520 --> 00:20:24.797 its power. Um do you have any messages for Europe 00:20:24.880 --> 00:20:27.597 related to this? >> Yeah, I mean I think so there are two 00:20:27.680 --> 00:20:29.357 sets of things, right? one is you know 00:20:29.440 --> 00:20:31.917 here we are in in Switzerland and I look 00:20:32.000 --> 00:20:34.877 at uh the the pharma or the financial 00:20:34.960 --> 00:20:38.317 sector you know OBC sw they're uh they 00:20:38.400 --> 00:20:42.077 they do do a big job in in in this country as in in Europe but they're also 00:20:42.160 --> 00:20:45.757 international brands and international operations. So one thing that uh 00:20:45.840 --> 00:20:49.917 whenever I think about Europe is the Europeans are producing products and 00:20:50.000 --> 00:20:51.917 services that actually are going 00:20:52.000 --> 00:20:54.397 everywhere in the world and so therefore 00:20:54.480 --> 00:20:57.037 uh European competitiveness is about the 00:20:57.120 --> 00:21:01.437 competitiveness of their output globally not just in Europe. I think sometimes 00:21:01.520 --> 00:21:03.037 when you come to Europe there's a lot of 00:21:03.120 --> 00:21:05.677 conversation about just Europe. Uh but 00:21:05.760 --> 00:21:09.597 European economy is thrives and has you 00:21:09.680 --> 00:21:13.277 know thrived in the last whatever 200 years 300 years. the miracle of the west 00:21:13.360 --> 00:21:17.117 is fundamentally because of what has happened in Europe uh is because they 00:21:17.200 --> 00:21:20.637 were able to produce things that the world needed and so I would say that's 00:21:20.720 --> 00:21:22.797 number one and in order to do that you 00:21:22.880 --> 00:21:24.957 again I go back to the human capital 00:21:25.040 --> 00:21:27.757 here is just fantastic and world class 00:21:27.840 --> 00:21:30.717 uh you have to absolutely invest in uh 00:21:30.800 --> 00:21:35.437 producing uh you know having the energy and the tokens here which again you're 00:21:35.520 --> 00:21:39.277 attracting like as I said we're investing and others are investing uh 00:21:39.360 --> 00:21:43.357 the the data centers here so the question is what's that next generation 00:21:43.440 --> 00:21:47.357 of output that comes from here. Right. I always think about the German middle 00:21:47.440 --> 00:21:51.277 star whenever I go to a jeweler or a dentist in the United States. I'm 00:21:51.360 --> 00:21:53.677 surrounded by German middle. Right. 00:21:53.760 --> 00:21:56.717 Totally. >> Um it's just unbelievable engineering 00:21:56.800 --> 00:22:00.157 provice of that country. uh and now the 00:22:00.240 --> 00:22:01.917 question and by the way that's the point 00:22:02.000 --> 00:22:04.477 that they are producing industrial 00:22:04.560 --> 00:22:08.237 products which today are built in into 00:22:08.320 --> 00:22:11.117 it all the intelligence as well that 00:22:11.200 --> 00:22:13.517 data right so I know whenever we come to 00:22:13.600 --> 00:22:15.597 Europe everyone's like talking about 00:22:15.680 --> 00:22:18.317 sovereignty and data this data that 00:22:18.400 --> 00:22:21.117 guess what Europe actually should be 00:22:21.200 --> 00:22:25.037 much more concerned about access to 00:22:25.120 --> 00:22:27.357 their industrial companies their 00:22:27.440 --> 00:22:30.157 financial services companies of data 00:22:30.240 --> 00:22:35.517 from US and the rest of the world. Uh as opposed to just thinking that somehow by 00:22:35.600 --> 00:22:38.637 protecting Europe you're going to be competitive. You are only going to be 00:22:38.720 --> 00:22:43.917 competitive if the products coming out of Europe are globally competitive, 00:22:44.000 --> 00:22:47.677 right? >> Um and so that's I think what needs to 00:22:47.760 --> 00:22:49.757 change. Uh you know Europe has led in 00:22:49.840 --> 00:22:52.717 privacy that's fantastic has led in many 00:22:52.800 --> 00:22:57.197 aspects of even safety around AI and what have you. And that's a feature uh 00:22:57.280 --> 00:23:01.437 that's great. But you also have to complement it with by building locally 00:23:01.520 --> 00:23:03.837 and then also thinking globally what's 00:23:03.920 --> 00:23:06.397 the contribution this continent will 00:23:06.480 --> 00:23:08.717 make uh to the rest of the world which 00:23:08.800 --> 00:23:11.117 it has historically been a leader 00:23:11.200 --> 00:23:13.997 >> a leader. So do you think the whole idea 00:23:14.080 --> 00:23:17.757 around sovereignty of data is that being misunderstood? 00:23:17.840 --> 00:23:20.157 I I think that that when people talk 00:23:20.240 --> 00:23:25.357 about sovereignty first of all it's very important clearly 00:23:25.440 --> 00:23:28.237 and who wants to >> and in a week like this it's more 00:23:28.320 --> 00:23:29.357 important 00:23:29.440 --> 00:23:35.357 >> uh but that said uh it is you have to 00:23:35.440 --> 00:23:39.117 kind of think about how what is sovereignty mean like for example in the 00:23:39.200 --> 00:23:41.837 AI era the topic that's least talked 00:23:41.920 --> 00:23:44.237 about but I feel will be most talked 00:23:44.320 --> 00:23:47.677 about in in this uh this calendar will 00:23:47.760 --> 00:23:50.157 be the sovereignty of a firm. Just 00:23:50.240 --> 00:23:53.677 imagine if your firm you're not able to 00:23:53.760 --> 00:23:57.597 embed the tacit knowledge of the firm in 00:23:57.680 --> 00:24:01.277 a set of weights in a model that you 00:24:01.360 --> 00:24:03.997 control. By definition, you have no 00:24:04.080 --> 00:24:06.397 sovereignty. That means you're leaking 00:24:06.480 --> 00:24:11.197 enterprise value through some model company somewhere. In fact, that it's 00:24:11.280 --> 00:24:14.157 sort of fascinating that nobody's talking about that, right? It's like 00:24:14.240 --> 00:24:15.597 everybody's talking about everything 00:24:15.680 --> 00:24:18.717 else that is sort of you know outside of 00:24:18.800 --> 00:24:21.677 that whereas the most important thing is 00:24:21.760 --> 00:24:25.757 it really doesn't matter if you have in fact the data center where it runs is 00:24:25.840 --> 00:24:29.277 the least important thing quite honestly but like even there first of all the 00:24:29.360 --> 00:24:32.877 data centers all are all over just because speed of light is a real 00:24:32.960 --> 00:24:37.517 constraint uh and so therefore the data centers will be spread yes uh you will 00:24:37.600 --> 00:24:40.877 have digital you'll be able to encrypt everything you'll be able to have the 00:24:40.960 --> 00:24:43.197 keys with you all of these are much more 00:24:43.280 --> 00:24:45.677 techn technically solve problems but the 00:24:45.760 --> 00:24:48.877 one problem that will only be solved is 00:24:48.960 --> 00:24:51.597 by you having much more sovereignty over 00:24:51.680 --> 00:24:54.317 the you know tacid knowledge uh and 00:24:54.400 --> 00:24:56.157 control over the models and it's not a 00:24:56.240 --> 00:24:59.197 one-way enterprise value transfer um and 00:24:59.280 --> 00:25:01.757 so to me I think sovereignty requires 00:25:01.840 --> 00:25:05.037 real thought on what is it um you know 00:25:05.120 --> 00:25:08.237 control of destiny means that your your 00:25:08.320 --> 00:25:10.477 ability to produce something that is 00:25:10.560 --> 00:25:13.597 unique is preserved uh David Ricard was 00:25:13.680 --> 00:25:15.757 not wrong. There's comparative advantage 00:25:15.840 --> 00:25:18.237 in countries. Uh there is comparative 00:25:18.320 --> 00:25:21.357 advantage in firms that needs to be 00:25:21.440 --> 00:25:25.597 preserved even in the AI era. That's what'll give you real sovereignty. 00:25:25.680 --> 00:25:29.037 >> One last question. I know we're running out of time. 00:25:29.120 --> 00:25:33.117 Um in 5 years or 10 years, is there going to be one dominant model that 00:25:33.200 --> 00:25:37.677 we're all going to be using or are and how is Microsoft preparing for this? Are 00:25:37.760 --> 00:25:39.357 you going to be are we going to be using 00:25:39.440 --> 00:25:42.317 one model for for enterprise, one model 00:25:42.400 --> 00:25:44.957 for other other traits? 00:25:45.040 --> 00:25:47.197 >> You know, even in the last whatever 3 00:25:47.280 --> 00:25:49.357 years, four years that we've been at it, 00:25:49.440 --> 00:25:53.597 um the the reality at this point is it's 00:25:53.680 --> 00:25:57.117 a multimodel world, right? I mean the in 00:25:57.200 --> 00:26:01.437 fact if you think about it the in both 00:26:01.520 --> 00:26:05.357 there are going to be multiple models 00:26:05.440 --> 00:26:08.637 and the trick is really how do you take 00:26:08.720 --> 00:26:11.517 advantage of these multiple models and 00:26:11.600 --> 00:26:13.997 in fact build your own model by 00:26:14.080 --> 00:26:16.877 distilling these right uh so think of 00:26:16.960 --> 00:26:20.077 these models uh that you orchestrate to 00:26:20.160 --> 00:26:21.757 build your own model and more 00:26:21.840 --> 00:26:25.037 importantly you do what is described as 00:26:25.120 --> 00:26:27.837 orchestration or harness engineering. So 00:26:27.920 --> 00:26:31.917 the IP of any application or any firm is 00:26:32.000 --> 00:26:35.197 how do you use all these models with 00:26:35.280 --> 00:26:37.277 context engineering or your data? Yes. 00:26:37.360 --> 00:26:41.517 >> Right. So it's that three parts. So can I bring in all the models by the way uh 00:26:41.600 --> 00:26:44.077 which is close source, open source, 00:26:44.160 --> 00:26:47.837 build my own model, orchestrate them and 00:26:47.920 --> 00:26:51.437 feed it my data to change the trajectory 00:26:51.520 --> 00:26:54.157 of some outcome that I care about. 00:26:54.240 --> 00:26:56.637 That's it. That's the entire picture. So 00:26:56.720 --> 00:26:58.797 you can do it in like oh I produce a 00:26:58.880 --> 00:27:01.197 particular product or service. Uh first 00:27:01.280 --> 00:27:05.997 I got to do better better job in sales or better job in R&D or better job in 00:27:06.080 --> 00:27:09.917 finance or what have you. And you take that outcome and then you say can I use 00:27:10.000 --> 00:27:12.477 all the models orchestrate them and feed 00:27:12.560 --> 00:27:15.037 it my context and then in as a result of 00:27:15.120 --> 00:27:17.997 it the reasoning traces are really 00:27:18.080 --> 00:27:20.877 leading to some capability and models 00:27:20.960 --> 00:27:23.837 that I control as my IP. As long as 00:27:23.920 --> 00:27:28.157 firms can answer that question, they're going to be getting ahead. 00:27:28.240 --> 00:27:30.637 >> Ladies and gentlemen, let's uh thank 00:27:30.720 --> 00:27:37.400 Satya, my friend. Thank you [applause] for 00:27:38.160 --> 00:27:41.997 and hopefully this is the beginning of many great dialogue and conversations 00:27:42.080 --> 00:27:45.117 here at the World Economic Forum. Thank you everyone. 00:27:45.200 --> 00:27:48.360 >> Thank you.