WEBVTT - Gary Cohn talks AI and How It Productive It Is 0:00:02.520 --> 0:00:08.520 Bloomberg Audio Studios, podcasts, radio news bringing out to that 0:00:08.600 --> 0:00:11.920 special conversation that we mentioned at Bloomberg in Vest, Gary Coheneweischer, 0:00:11.960 --> 0:00:13.840 IBM and Bloomberg's Matt Miller. 0:00:13.920 --> 0:00:16.280 Let's watch and listen live. Yeah, we have right now. 0:00:17.000 --> 0:00:20.120 During that period of time, you had to pre program 0:00:20.160 --> 0:00:23.840 a huge database with all of the answers because we 0:00:23.960 --> 0:00:28.640 didn't have the real time connectivity to the data around 0:00:28.640 --> 0:00:30.760 the world. Today, if you were going to play Jeopardy, 0:00:30.760 --> 0:00:33.120 you would completely do it differently. You would just allow 0:00:33.640 --> 0:00:37.680 the AI machine to go get the answer instantaneously. Forty 0:00:37.760 --> 0:00:39.479 years ago, you had to sort of put the answers 0:00:39.479 --> 0:00:42.120 in the room and hope that Jeopardy asked the questions 0:00:42.159 --> 0:00:45.560 in the categories you wanted. It was easier to program 0:00:45.560 --> 0:00:48.720 all the chess moves. Today, you know, AI has evolved 0:00:48.760 --> 0:00:51.880 dramatically and I think we're all living through this. I 0:00:51.920 --> 0:00:54.840 would say evolution. I'm not sure it's a revolution evolution 0:00:54.920 --> 0:00:58.120 of AI. You know, we've gone from the concept and 0:00:58.160 --> 0:01:00.760 if you look at sort of CEO surveys or corporate surveys, 0:01:01.040 --> 0:01:02.800 you know, a year or two ago, I think it 0:01:02.880 --> 0:01:06.920 was ninety about ninety percent of CEOs said, they, you know, 0:01:07.160 --> 0:01:10.000 clearly we're going to be involved in use AIAI. And 0:01:10.080 --> 0:01:12.640 of those ninety percent, less than ten percent could actually 0:01:12.680 --> 0:01:13.479 tell you what they were going. 0:01:13.480 --> 0:01:14.080 To use it for. 0:01:14.440 --> 0:01:17.360 They just felt like they needed to talk about using it. Today, 0:01:17.640 --> 0:01:21.039 I think the amount of CEOs they are using AI 0:01:21.520 --> 0:01:24.520 continues to grow, but more importantly, they actually know what 0:01:24.560 --> 0:01:30.199 they're going to use it for, and the definitional outcomes 0:01:30.280 --> 0:01:34.479 of where AI can enhance your business is becoming clear. 0:01:36.040 --> 0:01:37.840 That's where they are today. I think it's going to 0:01:37.880 --> 0:01:41.560 continue to evolve to where you can use AI in 0:01:41.600 --> 0:01:45.520 your business and where the paybacks are and how you create. 0:01:45.280 --> 0:01:47.280 Real productivity growth. 0:01:48.200 --> 0:01:50.960 I won't geek out on productivity growth, but at the 0:01:51.040 --> 0:01:53.080 end of the day, you know, if you want to 0:01:53.080 --> 0:01:55.640 grow GDP and you want to grow your economy, it 0:01:55.680 --> 0:01:59.600 is all about getting more productivity, getting more units of 0:01:59.640 --> 0:02:00.920 work your labor force. 0:02:01.080 --> 0:02:05.320 So I think about this a lot in a competitive sense. 0:02:05.440 --> 0:02:07.720 If you're a big business and you haven't figured out 0:02:07.760 --> 0:02:08.280 how you're going to. 0:02:08.360 --> 0:02:12.160 Use AI yet, are you too late? Have you lost out? 0:02:12.840 --> 0:02:15.640 You're never too late. You're never too late. You know 0:02:15.919 --> 0:02:18.680 what you may be late is you may be at 0:02:18.720 --> 0:02:22.880 a competitive disadvantage to your nearest competitor because they may 0:02:22.919 --> 0:02:25.399 be able to do certain things more efficiently than you, 0:02:25.919 --> 0:02:28.960 But you can catch up quickly, so you're not that 0:02:29.120 --> 0:02:31.760 far behind. So you have to sort of figure out 0:02:31.800 --> 0:02:34.120 where the opportunities are in your business. And it depends 0:02:34.160 --> 0:02:37.160 on size, scope scale of your business, so like a 0:02:38.800 --> 0:02:41.400 very obvious place where in a bigger company. And this 0:02:41.440 --> 0:02:43.720 is something we did at IBM, and we've talked about it. 0:02:43.760 --> 0:02:46.720 We like being client zero for all of our products. 0:02:47.240 --> 0:02:50.120 You know, we did something relatively simple. It sounds simple, 0:02:50.160 --> 0:02:52.919 but it was not that simple, an HR chat pot. 0:02:53.440 --> 0:02:56.000 You think of what most people use. 0:02:56.080 --> 0:03:00.240 HR for after the day one hiring experience of what 0:03:00.280 --> 0:03:02.480 they want is, you know, they go buy an apartment, 0:03:02.520 --> 0:03:04.600 they go rent an apartment, they get some they need 0:03:04.639 --> 0:03:06.520 a reference letter, they need a reference leative. What day 0:03:06.520 --> 0:03:08.280 did I start working? How long have I worked here? 0:03:08.440 --> 0:03:10.680 What was my last year's conversation? What was my bonus? 0:03:11.000 --> 0:03:14.240 And you want that scent like that minute. And you 0:03:14.280 --> 0:03:16.919 know in a year, a year or two ago, used 0:03:16.919 --> 0:03:19.359 to call someone in HR, used to give them the information. 0:03:19.440 --> 0:03:20.960 You said, tell them where you want it sent, and 0:03:21.000 --> 0:03:23.200 you hope and pray they send it out quickly. Today 0:03:23.480 --> 0:03:26.560 you're at IBM, you will type it into a HR 0:03:26.639 --> 0:03:29.600 chatbot and in a minute you said hit send, the 0:03:29.720 --> 0:03:32.600 letter will go out. So people are actually very happy. 0:03:32.600 --> 0:03:34.040 Oh my god, I get what I want. I get 0:03:34.080 --> 0:03:36.160 all this information. You have information about your benefit and 0:03:36.720 --> 0:03:39.720 information about your healthcare. We do it all real time. 0:03:39.800 --> 0:03:43.160 We cut six hundred people out of our HR department. 0:03:43.160 --> 0:03:45.960 Now we didn't fire them, we redeployed them to places 0:03:46.000 --> 0:03:47.280 where they can be more useful. 0:03:47.680 --> 0:03:49.120 And our satisfaction scores for. 0:03:49.640 --> 0:03:53.200 Our employees using HR has gone up dramatically because they 0:03:53.200 --> 0:03:56.560 get real time answers. So it depends what you're trying 0:03:56.600 --> 0:03:58.840 to achieve in your company. You know, there's a lot 0:03:58.880 --> 0:04:02.160 of people using AI today and. 0:04:02.160 --> 0:04:04.320 Code assist in writing new code. 0:04:04.440 --> 0:04:06.240 So if you're in a big code writing shop or 0:04:06.280 --> 0:04:09.240 writing new code is important to you, and you're at 0:04:09.240 --> 0:04:10.920 a disadvantage. 0:04:10.160 --> 0:04:12.000 Because your competitors are using a. 0:04:13.640 --> 0:04:17.120 Code assist program and you're you're at a competitive disadvantage. 0:04:17.240 --> 0:04:19.800 You take a lot of people here on financial services industry, 0:04:19.880 --> 0:04:22.919 most banks happen to know what banks do. Most banks 0:04:23.040 --> 0:04:27.800 historically build a technology infrastructure on a sort of as 0:04:27.880 --> 0:04:30.680 needed basis, and if you go pull back the covers. 0:04:30.680 --> 0:04:33.800 They've got lots of different systems in a variety of 0:04:33.839 --> 0:04:36.680 different languages, and when you try and integrate them and 0:04:36.680 --> 0:04:38.960 have them talk to each other, it's very difficult. So 0:04:39.800 --> 0:04:41.960 every bank during the course of the last decade or 0:04:42.000 --> 0:04:44.520 twenty years tried to think, how do I get all 0:04:44.720 --> 0:04:46.800 of my software? How do I get all my systems 0:04:46.839 --> 0:04:50.080 on a common language. It was an arduous task that 0:04:50.240 --> 0:04:53.920 was cost prohibitive. Today it's not cost prohibitive. Today you 0:04:54.000 --> 0:04:58.360 can get a code assist bot to basically port one 0:04:58.520 --> 0:05:01.760 language over to another pro another language, and it's pretty 0:05:01.760 --> 0:05:03.919 flawless and how it happens. So there's a lot of 0:05:03.960 --> 0:05:08.320 places evolving today where AI is very useful, and it. 0:05:08.279 --> 0:05:10.720 Depends on your I think your circumstances. 0:05:10.720 --> 0:05:12.680 I want to ask you about the cost because I 0:05:12.680 --> 0:05:15.080 think the reason it was so daunting in my mind, 0:05:15.160 --> 0:05:19.479 especially for a information technology company, is I was looking 0:05:19.520 --> 0:05:22.280 at open AI and thinking, you've got to invest one 0:05:22.320 --> 0:05:26.159 hundred billion dollars to create a large language model, and 0:05:26.200 --> 0:05:29.359 only a few of these giant hyperscalers in Silicon Valley 0:05:29.360 --> 0:05:29.840 can do it. 0:05:30.000 --> 0:05:30.200 Now. 0:05:30.240 --> 0:05:33.600 Obviously, Deep Seek, although there are questions about how exactly 0:05:33.640 --> 0:05:35.440 they did it how much it costs, shows that you 0:05:35.440 --> 0:05:36.520 can do it for far less. 0:05:37.279 --> 0:05:39.839 Well, thank you the Chinese Deep Seek for proving out 0:05:39.839 --> 0:05:41.320 what a lot of us knew in this country but 0:05:41.360 --> 0:05:43.720 no one wanted to listen to. So if you look 0:05:43.720 --> 0:05:45.240 at what we've been doing at IBM, you look at 0:05:45.279 --> 0:05:46.880 what met has been doing, you look at a lot 0:05:46.920 --> 0:05:49.719 of companies here in the United States been doing. We've 0:05:48.960 --> 0:05:51.120 been building small models. 0:05:51.880 --> 0:05:54.480 What we have found and Deep Seek sort of proved 0:05:54.520 --> 0:05:56.640 this out. The small models are much. 0:05:56.480 --> 0:06:00.680 More efficient, they compute faster, they take less power, but 0:06:00.760 --> 0:06:03.839 they're not a model that can do everything. They're sort 0:06:03.880 --> 0:06:07.200 of targeted models. So the things that I was talking 0:06:07.240 --> 0:06:09.799 to you about, we have small models that are used 0:06:09.800 --> 0:06:14.679 for specific targeted goals. I think that is the future. 0:06:14.720 --> 0:06:17.120 We think that is the future of AI. We think 0:06:17.400 --> 0:06:21.040 most companies will engage with as many small models as 0:06:21.040 --> 0:06:23.640 they need to get done what they need to get 0:06:23.640 --> 0:06:26.440 done where they're using AI, and it will be vastly 0:06:26.480 --> 0:06:29.520 more efficient, vastly more cost effective for them. And I 0:06:29.520 --> 0:06:31.719 think this is the future where we're going. Look, will 0:06:31.720 --> 0:06:35.440 there be some large language models out there that are effective, Yes, 0:06:35.600 --> 0:06:38.120 but they will sort of be the catch all models 0:06:38.160 --> 0:06:39.360 for everything, so. 0:06:40.920 --> 0:06:43.080 It still costs a lot of money to invest. I 0:06:43.080 --> 0:06:47.480 think Moody's estimated it's going to be two billion dollars 0:06:47.520 --> 0:06:49.559 in investment over the next five years, which doesn't seem 0:06:49.560 --> 0:06:51.520 that far off when I look at the kapex plans 0:06:51.920 --> 0:06:56.039 of these mag seven companies. You and I both lived 0:06:56.080 --> 0:06:59.320 through the Internet bubble and watched so many companies put 0:06:59.360 --> 0:07:03.640 money into paying down fiber cable and build the infrastructure 0:07:03.640 --> 0:07:07.000 the data centers of Web one point zero for a 0:07:07.040 --> 0:07:09.680 lot of CAPEX when they weren't getting a lot of 0:07:09.720 --> 0:07:11.960 return on investment, at least not for years if they 0:07:12.000 --> 0:07:12.559 ever did. 0:07:13.120 --> 0:07:15.840 Do you see real parallels here? Yeah? Yes. 0:07:16.400 --> 0:07:23.120 Look, as you're building technological infrastructure and backbone, you're building 0:07:23.160 --> 0:07:26.960 it based on the available what's available today. 0:07:27.000 --> 0:07:29.680 In what you know today, the whole. 0:07:31.000 --> 0:07:38.280 AI backbone infrastructure, compute technology, it's moving very fast. But 0:07:38.400 --> 0:07:41.360 you can't wait for version two point zero or version 0:07:41.360 --> 0:07:43.120 three point zero, version four point oh, you have to 0:07:43.120 --> 0:07:46.360 build based on what's build what's available today. What's available 0:07:46.360 --> 0:07:49.560 today will be obsolete a lot of it before it 0:07:49.560 --> 0:07:52.240 gets built, and it will get replaced. 0:07:51.760 --> 0:07:53.679 By newer and newer versions. 0:07:53.680 --> 0:07:56.320 So there is a bunch of money being spent today, 0:07:56.440 --> 0:07:59.880 they'll have very low returns to no returns. But if 0:07:59.920 --> 0:08:02.960 you don't start investing, you'll never start. 0:08:03.040 --> 0:08:04.360 So you sort of have this. 0:08:04.320 --> 0:08:07.280 Prisoner's dilemma that I have to start building this data 0:08:07.320 --> 0:08:09.600 center and I think I'm going to be putting x 0:08:09.760 --> 0:08:11.880 y z chip in, but by the way it gets 0:08:11.920 --> 0:08:14.600 built in two years, I may not be putting most 0:08:14.720 --> 0:08:17.080 likely will not be putting that chip in, And the 0:08:17.120 --> 0:08:19.280 amount of power I think I need to run that 0:08:19.360 --> 0:08:22.840 data center may be dramatically less because the next chip 0:08:23.360 --> 0:08:25.760 I bet is going to be a lot more efficient 0:08:25.960 --> 0:08:28.240 than the chip we have today. So a lot of 0:08:28.280 --> 0:08:32.160 this build you can't wait for the efficient frontier because someone's. 0:08:31.840 --> 0:08:32.959 Always going to be out there. 0:08:33.160 --> 0:08:35.120 You have to keep building, and there will be a 0:08:35.120 --> 0:08:37.920 lot of waste to get there. I just think it's inevitable. 0:08:38.040 --> 0:08:40.360 And then you know, these AI models themselves are going 0:08:40.440 --> 0:08:42.800 to build upon themselves, and they're gonna get smarter and smarter, 0:08:43.040 --> 0:08:46.079 and they're going to optimize and re optimize, and everything's 0:08:46.120 --> 0:08:49.240 going to get better and faster and quicker, and ultimately 0:08:49.280 --> 0:08:52.120 the AI of this of the data center is going 0:08:52.160 --> 0:08:53.679 to tell you how to build better data center, how 0:08:53.679 --> 0:08:57.079 to build better connectivity, how to optimize the optimization. 0:08:57.480 --> 0:08:59.120 What do you do though as an investor, I mean, 0:08:59.800 --> 0:09:02.319 you know, as an IBM or as a Microsoft or 0:09:02.360 --> 0:09:04.720 as you point out, they don't have any choice, They've 0:09:04.720 --> 0:09:07.719 got to build now. But what do you do as 0:09:07.760 --> 0:09:09.280 a as a private investor or what do you do 0:09:09.280 --> 0:09:10.520 as someone who's looking at markets? 0:09:10.520 --> 0:09:13.280 How do you analyze that? Well, it's interesting if you 0:09:13.320 --> 0:09:13.760 look at the. 0:09:15.640 --> 0:09:18.880 History and the history of technology is not that old, 0:09:19.200 --> 0:09:21.640 and if you look at you know, sort of what's 0:09:21.679 --> 0:09:27.560 going on here. Ultimately, the bigger companies are going to have. 0:09:29.040 --> 0:09:31.480 A real big impact in this now. 0:09:31.800 --> 0:09:33.880 It's really and then I'll point out that all the 0:09:33.920 --> 0:09:36.680 bigger companies that we talk about today, whether it be 0:09:36.760 --> 0:09:40.360 the Metas or the Googles. You know, I remember when 0:09:40.360 --> 0:09:42.120 Facebook started, it was a really small company. 0:09:42.320 --> 0:09:43.880 And I'm not that old. I'm old enough, but I'm 0:09:43.920 --> 0:09:44.400 not that old. 0:09:44.400 --> 0:09:46.880 So all these big companies, I remind people, start out 0:09:46.880 --> 0:09:49.560 as really small companies. So there are a bunch of 0:09:49.600 --> 0:09:52.440 small companies so that are going to be really big 0:09:52.440 --> 0:09:56.559 companies in less in that path, one of these bigger 0:09:56.559 --> 0:09:58.679 companies can buy them out. You know, We've now got 0:09:58.679 --> 0:10:00.600 companies that can go spend on hundred million dollars in 0:10:00.679 --> 0:10:04.240 buy out a company you know, relatively easily today. So 0:10:05.760 --> 0:10:08.880 if you were looking at what you think is the 0:10:08.920 --> 0:10:12.440 most likely path, I think the most likely path is 0:10:12.480 --> 0:10:14.680 the big I don't know if it's the mag seven, 0:10:14.720 --> 0:10:18.800 but the bigger companies who are spending the most in 0:10:18.920 --> 0:10:24.320 CAPEX are the most likely winners, and they will you know, 0:10:24.360 --> 0:10:25.559 they many will. 0:10:25.400 --> 0:10:26.680 Fund some of these startups. 0:10:27.600 --> 0:10:30.000 And almost all of the big companies have huge venture 0:10:30.080 --> 0:10:32.640 arms because they're trying to spend money in these startups. 0:10:32.679 --> 0:10:33.760 They're trying to get. 0:10:33.679 --> 0:10:37.360 Opportunities to let creative people create, and then as those 0:10:37.400 --> 0:10:39.800 products become really valuable, they're just going to buy them 0:10:39.840 --> 0:10:41.960 and merge them into the bigger companies. So, you know, 0:10:42.280 --> 0:10:44.880 I guess my investment advice is, like, you can own 0:10:44.920 --> 0:10:46.959 some of these small companies and most are private, and 0:10:46.960 --> 0:10:50.200 they're probably interesting to own a portfolio of them. But 0:10:50.280 --> 0:10:52.200 if you ask me where I think this is going 0:10:52.240 --> 0:10:54.000 to be a decade from now, I think the bigger 0:10:54.040 --> 0:10:55.520 companies are going to dominate the space. 0:10:55.640 --> 0:10:58.120 Is that part of what you're doing at IBM? I mean, 0:10:58.200 --> 0:11:02.240 do you help Arvin Krishna maybe zone in on some targets. 0:11:02.559 --> 0:11:05.040 We definitely have a ventures portfolio. 0:11:05.360 --> 0:11:09.040 We're always looking for something that's on the cutting edge 0:11:09.160 --> 0:11:11.160 that fits within our parameters. 0:11:11.440 --> 0:11:12.080 You know, where there's. 0:11:12.000 --> 0:11:15.880 Hybrid cloud, open AI, quantum, you know. 0:11:15.960 --> 0:11:17.960 But we're very much in the open source. 0:11:18.000 --> 0:11:20.160 So we're gonna buy open source company if it fits 0:11:20.200 --> 0:11:23.880 within our parameters and we think it's got an interesting opportunity. 0:11:24.320 --> 0:11:27.160 It's in our shareholder's best interest for us to get 0:11:27.200 --> 0:11:29.000 involved in the company early, not late. 0:11:29.400 --> 0:11:33.040 By the way, a little bit detached from the specific 0:11:33.120 --> 0:11:35.560 AI conversation, what do you think about the M and 0:11:35.600 --> 0:11:37.640 I M and A landscape right now? 0:11:37.679 --> 0:11:38.760 When we came into. 0:11:38.480 --> 0:11:42.360 This year, me and Shanali and Katie, we're joking about 0:11:42.400 --> 0:11:44.480 an opinion writer who said, maybe there won't be that 0:11:44.559 --> 0:11:47.360 many deals because everybody said this is going to be 0:11:47.400 --> 0:11:50.080 the year you know that the regulators drop their guard 0:11:50.120 --> 0:11:53.680 and everybody can get together. Now it's starting to look 0:11:53.720 --> 0:11:54.880 like that may not be the case. 0:11:55.160 --> 0:11:59.800 Well, that was the total market euphoria from November election 0:12:00.200 --> 0:12:06.720 January twentieth. Was deregulation, ease of doing business, the FDC, 0:12:06.880 --> 0:12:10.800 the CFDC, the SEC, everything C is going to get 0:12:10.800 --> 0:12:14.160 easier and be much more pro business and we're going 0:12:14.200 --> 0:12:18.480 to have this huge influx, this huge, massive explosion of 0:12:18.559 --> 0:12:21.000 deals of M and A, deals of IPO deals. 0:12:21.920 --> 0:12:24.880 Look, we're forty five or forty six days into this. 0:12:25.400 --> 0:12:27.360 I don't want to judge this book by its cover, 0:12:27.880 --> 0:12:30.600 but the first forty five days we've not had an 0:12:30.640 --> 0:12:34.000 explosion of deals. In fact, probably based on what's going 0:12:34.040 --> 0:12:35.400 on in markets right now, in market. 0:12:35.200 --> 0:12:37.280 Multiples, we've probably lost deals. 0:12:37.440 --> 0:12:40.880 So I would say the calendar from January twenty today 0:12:41.040 --> 0:12:44.200 is smaller than it was. That's not a regulatory statement, 0:12:44.200 --> 0:12:48.400 that's a valuation statement. And you also have to be 0:12:48.440 --> 0:12:51.960 reminded that some of these commissioners are not in place 0:12:52.440 --> 0:12:55.920 heads of SEC. Even the FDC doesn't have a full 0:12:56.000 --> 0:12:58.679 Republican slate, it's got two Republicans too Democrats, so it's 0:12:58.720 --> 0:13:02.000 hard to get things through on to too. These things 0:13:02.080 --> 0:13:05.640 take more time, So I think some of that euphoria 0:13:05.720 --> 0:13:09.120 is still there, but I think more people are taking 0:13:09.160 --> 0:13:12.000 the approach of Okay, you're going to have to show me. 0:13:12.360 --> 0:13:15.199 It's a different environment than just trust me. I think 0:13:15.200 --> 0:13:16.559 we've gone from trust me to show me. 0:13:17.280 --> 0:13:19.080 I want to get back to regulation in a minute, 0:13:19.360 --> 0:13:22.640 as it pretends to AI, But you mentioned productivity. You 0:13:22.640 --> 0:13:25.640 don't want to geek out on it. We haven't seen 0:13:25.800 --> 0:13:29.199 much of an impact yet from. 0:13:29.000 --> 0:13:30.640 A on productivity. 0:13:30.679 --> 0:13:33.520 I actually have a chatbot now open next to my 0:13:33.520 --> 0:13:36.720 Bloomberg screens every morning, so make you more productive. It's 0:13:36.840 --> 0:13:40.480 I'm actually getting a lot more detail into my newscast 0:13:40.559 --> 0:13:41.520 than I would otherwise. 0:13:43.080 --> 0:13:44.800 I use it as more of a glorified search. 0:13:44.880 --> 0:13:47.720 But we haven't really seen it boost the numbers economically, 0:13:48.559 --> 0:13:51.160 And I wonder when you think we're going to see 0:13:51.160 --> 0:13:53.120 that and if it's going to actually have an effect 0:13:53.200 --> 0:13:55.839 on jobs. You mentioned that you've moved HR people out 0:13:55.840 --> 0:13:59.520 because you can replace them with in that role with AI. 0:13:59.840 --> 0:14:02.480 Are we going to see you know, an uptick in 0:14:03.360 --> 0:14:06.400 or a loss in non farm payrolls because of it? 0:14:06.640 --> 0:14:10.520 So we've also created you know, artificial workers, you know, 0:14:10.559 --> 0:14:13.000 So we now can create artificial workers in certain in 0:14:13.000 --> 0:14:16.560 our consulting business. We now have artificial workers who you know, 0:14:16.679 --> 0:14:19.680 will work twenty four hour days, which is. 0:14:19.680 --> 0:14:22.560 Kind of nice, just like Doge. Yeah, so. 0:14:24.040 --> 0:14:25.720 I don't know if I'm going to make that you 0:14:25.800 --> 0:14:29.560 made that comparison, I will not own that comparison. So 0:14:29.680 --> 0:14:33.160 let let's talk about your question. I think your question is, 0:14:33.920 --> 0:14:36.840 you know, if AI is going to be successful, are 0:14:36.840 --> 0:14:39.720 we going to see a real impact to labor and 0:14:39.800 --> 0:14:41.760 employment numbers around the world? 0:14:41.880 --> 0:14:43.080 Let leonne in the United States. 0:14:43.680 --> 0:14:48.560 So my pretty bold answer is, and I always say 0:14:48.600 --> 0:14:50.880 this to rooms, to people like this who usually first 0:14:50.960 --> 0:14:54.000 question I get is isn't this devastating the economy? I say, look, 0:14:54.560 --> 0:14:56.240 I'll give you the rest of the time I'm up here, 0:14:56.240 --> 0:14:57.840 which is very short at this point, so it's not 0:14:57.880 --> 0:14:59.640 a lot of time. I'll give you the rest time 0:14:59.640 --> 0:15:03.200 I'm up to come up with a major technological advancement 0:15:03.640 --> 0:15:07.080 that destroyed jobs, and you can go. 0:15:07.040 --> 0:15:08.480 Back and you can take anyone you want. 0:15:08.640 --> 0:15:11.440 And so like, let's just take the internal combustion engine 0:15:11.880 --> 0:15:14.920 that was supposed to destroy jobs because all those people 0:15:14.920 --> 0:15:17.960 they cleaned stalls and put. 0:15:17.800 --> 0:15:20.400 Shoes on horses, they were going to lose their jobs. 0:15:20.800 --> 0:15:22.280 Well, but all of a sudden, the next day when 0:15:22.320 --> 0:15:26.040 we started putting internal combustion engines on the streets. We 0:15:26.080 --> 0:15:29.600 needed mechanics, we needed gas stations, we needed we needed 0:15:29.640 --> 0:15:33.640 different skill sets, different people. But all of a sudden, 0:15:33.640 --> 0:15:36.880 there were more jobs created because people could go you know, 0:15:36.960 --> 0:15:38.680 fifty miles a day or a hundred miles a day 0:15:38.720 --> 0:15:42.400 versus two miles a day, so their landscape, their business 0:15:42.440 --> 0:15:46.280 model expanded. It's no different than when the internet. Like 0:15:46.880 --> 0:15:49.560 I worked at a bank before we had email. We 0:15:49.640 --> 0:15:52.560 used to have every floor head had a person that 0:15:52.720 --> 0:15:55.320 printed the memos and put it in your mailbox and 0:15:55.360 --> 0:15:57.080 told you what time the meeting was or told you 0:15:57.080 --> 0:15:59.360 what time to go. What happened to that person who 0:15:59.400 --> 0:16:01.480 delivered the mail on the floor and printed the memos. 0:16:01.880 --> 0:16:04.520 You know, they sort of disappeared overnight. 0:16:05.280 --> 0:16:07.840 Now we didn't have less people. 0:16:07.840 --> 0:16:10.640 In fact, every bank, I guarantee you since the invention 0:16:10.680 --> 0:16:11.440 of email. 0:16:11.400 --> 0:16:12.720 Is double its size. 0:16:13.200 --> 0:16:16.480 Why because we got more productive. We got email coverage, 0:16:16.520 --> 0:16:18.400 so we could cover more clients. We could cover more 0:16:18.440 --> 0:16:20.880 people around the world. We could communicate quicker, we could 0:16:20.920 --> 0:16:24.720 communicate faster. I think AI is the next leg in 0:16:24.760 --> 0:16:29.160 this evolution of making us more productive. So look, will 0:16:29.200 --> 0:16:31.080 there be a change in skill sets? 0:16:31.200 --> 0:16:34.280 Yes? Will people do jobs that they hate? 0:16:34.720 --> 0:16:34.760 No. 0:16:35.440 --> 0:16:39.440 The machine can input data. The machine can take codes 0:16:39.480 --> 0:16:42.240 out of an operating room in a hospital and input 0:16:42.280 --> 0:16:45.400 them into the insurance reimbursement system. By the way, that 0:16:45.560 --> 0:16:47.560 job in a couple of hospitals has like forty fifty 0:16:47.560 --> 0:16:50.440 percent turnover rates because people hate doing that, they shouldn't 0:16:50.480 --> 0:16:52.640 do it. Those people in that hospital can now do 0:16:52.720 --> 0:16:54.840 a job where they try and work on recoveries and 0:16:54.880 --> 0:16:57.480 the hospital gets paid back. So I think you will 0:16:57.480 --> 0:17:01.200 see like this natural evolution of people out of jobs 0:17:01.240 --> 0:17:05.679 that they don't like, jobs with load dissatisfaction ratings, jobs 0:17:05.720 --> 0:17:08.399 that have higher satisfaction rates, that are more productive and 0:17:08.480 --> 0:17:09.400 grows the economy. 0:17:09.520 --> 0:17:11.160 That's the history of every. 0:17:11.040 --> 0:17:14.560 Other technological advancement we see. It's not overnight. That's why 0:17:14.600 --> 0:17:16.760 I call this an evolution, not a revolution. All right, 0:17:16.800 --> 0:17:19.280 and all this, by the way, we probably won't have time, 0:17:19.640 --> 0:17:22.840 but AI I believe is just a stepping stone to 0:17:22.960 --> 0:17:24.560 the real outcome, which is. 0:17:24.600 --> 0:17:28.800 Quantum, which is I'm guessing you also think ten to 0:17:28.840 --> 0:17:33.320 fifteen years off, No sooner, sooner, and how soon. 0:17:35.640 --> 0:17:37.960 Five years. 0:17:38.640 --> 0:17:42.119 The strides that is being made in quantum right now 0:17:43.040 --> 0:17:44.520 are large. 0:17:44.640 --> 0:17:45.600 I mean there's a lot more. 0:17:45.640 --> 0:17:47.480 Look again, there's a lot of private capital going in 0:17:47.480 --> 0:17:48.120 to quantum. 0:17:48.200 --> 0:17:51.760 It's big companies and small companies. It's all about error correction. 0:17:52.080 --> 0:17:55.000 How quickly you can correct their how quickly you can 0:17:55.040 --> 0:17:59.280 compute the data is pretty interesting right now. So you know, 0:17:59.520 --> 0:18:03.800 as we have evolve, you know, everything in this technology 0:18:04.119 --> 0:18:05.120 space is speeding up. 0:18:05.160 --> 0:18:07.520 So I'll you know, five years all right, fascinating. 0:18:07.720 --> 0:18:11.879 I wanted to ask about regulation because you mentioned AI 0:18:12.000 --> 0:18:14.040 is going to help us create better AI, and I 0:18:14.080 --> 0:18:15.320 wonder what else AI is. 0:18:15.280 --> 0:18:17.400 Going to decide to do for us. Do we need 0:18:17.440 --> 0:18:18.159 to regulate it? 0:18:18.240 --> 0:18:20.160 Do we need to have a heavy hand here, or 0:18:20.200 --> 0:18:22.840 is it okay to have a few South African billionaires 0:18:22.920 --> 0:18:26.320 just say like, go, I don't. 0:18:26.200 --> 0:18:27.320 Know, I don't know what you're talking about. 0:18:30.119 --> 0:18:32.239 I'll go to chat EPT and ask that what that is. 0:18:33.400 --> 0:18:35.840 So look, the answer is yes. 0:18:36.920 --> 0:18:39.080 And I sit on a couple of these big regulatory 0:18:39.080 --> 0:18:42.080 boards around the world, so I'll do this relatively quickly. 0:18:42.320 --> 0:18:46.560 If it's a regulated activity today, if you use a 0:18:47.200 --> 0:18:50.919 it should be a regulated activity using AI. So an 0:18:50.960 --> 0:18:53.320 example of that is if I'm going to call my 0:18:53.440 --> 0:18:56.440 doctor and tell my doctor that I got a scratchy throat, 0:18:56.480 --> 0:18:59.159 a headache, a swelling of this d D and he 0:18:59.320 --> 0:19:01.439 or she puts it in a machine and they're going 0:19:01.520 --> 0:19:05.080 to diagnose my disease and prescribe me medicine. That's a 0:19:05.200 --> 0:19:09.119 highly regulated business today. The AI machine that does that 0:19:09.520 --> 0:19:12.800 and the program that does should be highly regulated. On 0:19:12.880 --> 0:19:16.199 the flip side, if when I'm working out tomorrow morning, 0:19:16.520 --> 0:19:19.159 Spotify chooses a bad song for me and I have 0:19:19.200 --> 0:19:21.600 to hit skip, which they do a lot, you. 0:19:21.520 --> 0:19:23.800