1 00:00:05,120 --> 00:00:08,157 So Eric, when did you when did Google 2 00:00:08,240 --> 00:00:10,477 first start thinking about artificial 3 00:00:10,560 --> 00:00:12,157 intelligence? 4 00:00:12,240 --> 00:00:16,717 >> Larry Page was his PhD subject was AI. 5 00:00:16,800 --> 00:00:21,117 So even when Larry started the company, he wanted to do AI relatively 6 00:00:21,200 --> 00:00:23,517 interesting things. And the first real 7 00:00:23,600 --> 00:00:27,117 use of AI was we used AI to improve our 8 00:00:27,200 --> 00:00:29,757 ad system performance. This is before 9 00:00:29,840 --> 00:00:34,797 LMS and so forth and that worked. That was the first time I saw it used inside 10 00:00:34,880 --> 00:00:38,717 the business. >> Now you've written a book with on 11 00:00:38,800 --> 00:00:41,037 artificial intelligence. One of your 12 00:00:41,120 --> 00:00:45,757 co-authors was Henry Kissinger. Did Henry Kissinger really know anything 13 00:00:45,840 --> 00:00:49,997 about artificial intelligence? Why was he the co-author of that book? Well, he 14 00:00:50,080 --> 00:00:52,477 was uh he's obviously a brilliant man 15 00:00:52,560 --> 00:00:54,637 and he was very interested in it not for 16 00:00:54,720 --> 00:00:57,357 the technical reasons but because he had 17 00:00:57,440 --> 00:01:01,117 studied Kant when he was uh when he was 18 00:01:01,200 --> 00:01:03,917 the age of a college student well before 19 00:01:04,000 --> 00:01:07,117 any of us was born and when he saw AI 20 00:01:07,200 --> 00:01:11,677 when Demis presented to him he thought it would have an impact on human 21 00:01:11,760 --> 00:01:13,917 consciousness and he was very worried 22 00:01:14,000 --> 00:01:18,157 about it. Hey Joe. Um, usually when I'm 23 00:01:18,240 --> 00:01:21,277 interviewing people about artificial intelligence or something like that, 24 00:01:21,360 --> 00:01:25,197 there are people that have engineering backgrounds, their technology 25 00:01:25,280 --> 00:01:28,877 backgrounds. You're an accountant. So, what are you doing on a panel on 26 00:01:28,960 --> 00:01:30,957 artificial intelligence? 27 00:01:31,040 --> 00:01:34,157 >> Fortunately, of those 500,000 people at 28 00:01:34,240 --> 00:01:36,637 Deote, I have some good friends who are 29 00:01:36,720 --> 00:01:41,437 computer engineers and scientists and have deep expertise in the systems 30 00:01:41,520 --> 00:01:46,397 themselves. So, I I've got a little bit of help. How many people are devoted to 31 00:01:46,480 --> 00:01:49,917 artificial intelligence out of your 500,000 people? 32 00:01:50,000 --> 00:01:53,277 >> I would define every one of them is in some way leveraging the technology, 33 00:01:53,360 --> 00:01:57,197 helping clients use it. You can define a subset that obviously are the deep 34 00:01:57,280 --> 00:02:01,677 experts, but it's now pervasive across the entirety of the organization. 35 00:02:01,760 --> 00:02:05,677 >> Eric, if I want to make money on artificial intelligence, I'm an average 36 00:02:05,760 --> 00:02:10,157 investor. What is the best way for me to make money today on artificial 37 00:02:10,240 --> 00:02:12,477 intelligence? buying a big company like 38 00:02:12,560 --> 00:02:14,717 uh Alphabet or finding a smaller one 39 00:02:14,800 --> 00:02:17,357 that's just getting off the ground. 40 00:02:17,440 --> 00:02:19,837 >> Well, one way to understand the value 41 00:02:19,920 --> 00:02:23,037 that's being created in AI is that it 42 00:02:23,120 --> 00:02:24,957 will there will be businesses that will 43 00:02:25,040 --> 00:02:27,757 become much more successful because of 44 00:02:27,840 --> 00:02:30,957 AI. So, I think the best investment is 45 00:02:31,040 --> 00:02:35,197 to invest in the businesses where AI will make them transform. 46 00:02:35,280 --> 00:02:38,717 Um, it's probably pretty late to have all the gains that you have in the 47 00:02:38,800 --> 00:02:41,277 Magnificent 7, although it's not my area 48 00:02:41,360 --> 00:02:43,677 of expertise, simply because those are 49 00:02:43,760 --> 00:02:48,877 very highly valued. And the startups are now so highly valued when they get 50 00:02:48,960 --> 00:02:50,397 started that it's really hard to get 51 00:02:50,480 --> 00:02:53,437 into them. Um, if you really want to 52 00:02:53,520 --> 00:02:56,557 make money, it's actually easy. Found an 53 00:02:56,640 --> 00:03:00,077 agentic AI company. And I don't mean one 54 00:03:00,160 --> 00:03:02,317 designing agents. I mean build an agent 55 00:03:02,400 --> 00:03:07,517 to do something. This is the agentic period in AI and for the next year or 56 00:03:07,600 --> 00:03:09,277 two, everyone's going to build agents. 57 00:03:09,360 --> 00:03:12,797 The agents are all going to compete. If you build a better agent than anyone 58 00:03:12,880 --> 00:03:14,797 else, you can probably build one 59 00:03:14,880 --> 00:03:17,597 incredible company and own it yourself. 60 00:03:17,680 --> 00:03:21,437 >> Eric, um right now, um artificial 61 00:03:21,520 --> 00:03:23,997 intelligence is affecting adversely 62 00:03:24,080 --> 00:03:26,877 which um part of the economy the most 63 00:03:26,960 --> 00:03:29,117 and it was affecting positively which 64 00:03:29,200 --> 00:03:31,677 part of the economy the most. 65 00:03:31,760 --> 00:03:34,477 Uh well AI is at the moment contributing 66 00:03:34,560 --> 00:03:38,077 more than 1% to the GDP of America 67 00:03:38,160 --> 00:03:42,077 largely because of the data center buildout and I just started a data 68 00:03:42,160 --> 00:03:46,477 center company so I've been studying this the demands the the real demands 69 00:03:46,560 --> 00:03:48,477 from the hyperscalers the big companies 70 00:03:48,560 --> 00:03:51,197 Google and so forth are immense and when 71 00:03:51,280 --> 00:03:53,917 I talk to them they need 1 gawatt 5 72 00:03:54,000 --> 00:03:57,677 gawatt 10 gawatt each now the best study 73 00:03:57,760 --> 00:03:59,997 I've seen indicates that the industry in 74 00:04:00,080 --> 00:04:02,877 America needs 80 gawatt in the next 3 to 75 00:04:02,960 --> 00:04:06,317 5 years. Now 80 gawatt by the way let me 76 00:04:06,400 --> 00:04:08,717 tell you how big is that 77 00:04:08,800 --> 00:04:11,837 1.5 gawatt is the size of a nuclear 78 00:04:11,920 --> 00:04:14,477 power plant. So this is an enormous 79 00:04:14,560 --> 00:04:16,717 amount of energy. So that's the the 80 00:04:16,800 --> 00:04:19,837 economics right now are being most felt 81 00:04:19,920 --> 00:04:22,557 in the buildout of the infrastructure 82 00:04:22,640 --> 00:04:26,157 for the next wave of AI. Joe uh if you 83 00:04:26,240 --> 00:04:28,877 have 500,000 employees at Deote how many 84 00:04:28,960 --> 00:04:30,317 of them are going to be losing their 85 00:04:30,400 --> 00:04:33,437 jobs because of AI 86 00:04:33,520 --> 00:04:36,397 so with all of the adoption of 87 00:04:36,480 --> 00:04:40,157 technology that half a million today is 88 00:04:40,240 --> 00:04:44,157 greater than the number that we had a year ago will there be some displacement 89 00:04:44,240 --> 00:04:46,157 and shifting in certain functions that 90 00:04:46,240 --> 00:04:49,197 one did manually of course as with every 91 00:04:49,280 --> 00:04:51,277 wave of technology adoption but we're 92 00:04:51,360 --> 00:04:54,237 still a big believer that great people 93 00:04:54,320 --> 00:04:58,557 great professionals with domain expertise are actually the key to 94 00:04:58,640 --> 00:05:01,117 unlocking the technology and that as 95 00:05:01,200 --> 00:05:05,757 with every prior wave of technology change on a net basis you actually 96 00:05:05,840 --> 00:05:09,517 create more jobs than you disrupt that'll be tested but that's our base 97 00:05:09,600 --> 00:05:13,437 case >> okay so in your consulting business what 98 00:05:13,520 --> 00:05:17,197 has been some more significant the rise of the internet or the rise of 99 00:05:17,280 --> 00:05:21,277 artificial intelligence >> well if you look at the rise of the 100 00:05:21,360 --> 00:05:25,677 internet and the impact over a 30 plus year period in the acid to date that's 101 00:05:25,760 --> 00:05:27,917 more significant but in a confined time 102 00:05:28,000 --> 00:05:30,157 period in terms of this really thrusting 103 00:05:30,240 --> 00:05:35,837 onto the scene over the last 3 or so years it is a more accelerated impact 104 00:05:35,920 --> 00:05:39,997 than what we would have seen with the internet and right now I don't think 105 00:05:40,080 --> 00:05:42,717 anybody doubts the power of the 106 00:05:42,800 --> 00:05:45,437 technology the applicability across the 107 00:05:45,520 --> 00:05:48,157 enterprise every sector every business 108 00:05:48,240 --> 00:05:51,437 function the issue is it's really hard 109 00:05:51,520 --> 00:05:56,397 to actually bring it to life in a large complex established bureaucratic 110 00:05:56,480 --> 00:05:58,077 enterprise. You've got to actually be 111 00:05:58,160 --> 00:06:01,517 willing to drive a fundamental endtoend 112 00:06:01,600 --> 00:06:04,557 change in the business processes. And 113 00:06:04,640 --> 00:06:08,557 history tells us that there's a lot of stickiness in the system. And that 114 00:06:08,640 --> 00:06:11,517 diffusion from the the lab in Silicon 115 00:06:11,600 --> 00:06:14,317 Valley to the actual heartland to the 116 00:06:14,400 --> 00:06:19,357 factory, the office that people work in, it takes time. It takes effort. So we're 117 00:06:19,440 --> 00:06:22,957 in the middle of a a massive transformation that's going to take many 118 00:06:23,040 --> 00:06:23,917 years. 119 00:06:24,000 --> 00:06:27,037 >> Yeah, David, let go ahead. So this has 120 00:06:27,120 --> 00:06:30,317 always been true of technology adoption 121 00:06:30,400 --> 00:06:32,317 and the industry calls this the 122 00:06:32,400 --> 00:06:34,957 technology overhang. The industry we 123 00:06:35,040 --> 00:06:40,237 arrogantly say we keep producing things that you can't absorb fast enough. Of 124 00:06:40,320 --> 00:06:45,037 course, if we made it easier, maybe it would would on the jobs. There's 125 00:06:45,120 --> 00:06:49,917 beginning to be evidence of real job loss in two categories. One is in 126 00:06:50,000 --> 00:06:54,157 customer service, which is highly automated now, and the other one is in 127 00:06:54,240 --> 00:06:59,437 low-end or entry- level software programming, my field. Uh, and that 128 00:06:59,520 --> 00:07:00,957 shouldn't surprise you because the 129 00:07:01,040 --> 00:07:04,077 software programming is getting so good, 130 00:07:04,160 --> 00:07:08,317 you really can change your modality and you should expect that that will 131 00:07:08,400 --> 00:07:12,397 continue. The real question is when do other when do other parts of jobs get 132 00:07:12,480 --> 00:07:15,837 affected? >> Let's suppose I miss the internet uh 133 00:07:15,920 --> 00:07:20,797 wave and I miss the AI wave as an investor, but I'm always looking for the 134 00:07:20,880 --> 00:07:24,717 next big wave that I can get on the ground floor. What is the next big 135 00:07:24,800 --> 00:07:27,117 technology wave that I could get on the 136 00:07:27,200 --> 00:07:29,517 ground floor of? Is it is it quantum 137 00:07:29,600 --> 00:07:33,517 computing or some other thing uh like uh 138 00:07:33,600 --> 00:07:37,677 uh nuclear fusion? What would you say is the next big wave I should invest in if 139 00:07:37,760 --> 00:07:38,797 I want to be on the ground floor? 140 00:07:38,880 --> 00:07:42,397 >> So the next big wave is AI, right? We're 141 00:07:42,480 --> 00:07:46,157 not done and in fact the wave is just getting started. That's not I don't 142 00:07:46,240 --> 00:07:48,397 agree with you David. Um we're going 143 00:07:48,480 --> 00:07:50,477 from language to language which is think 144 00:07:50,560 --> 00:07:53,357 of it as chat GBT to language to action 145 00:07:53,440 --> 00:07:56,237 through through agents and if you spend 146 00:07:56,320 --> 00:07:59,037 any time with deep research or spend 147 00:07:59,120 --> 00:08:01,757 some time with open AAI R3 spend some 148 00:08:01,840 --> 00:08:05,117 time with DeepSeek V3 whatever the power 149 00:08:05,200 --> 00:08:07,597 of the reasoning models of these things 150 00:08:07,680 --> 00:08:10,877 is frightening right it's graduate level 151 00:08:10,960 --> 00:08:13,117 MIT Stanford reasoning right at your 152 00:08:13,200 --> 00:08:16,237 fingertips we're not done yet and indeed 153 00:08:16,320 --> 00:08:20,157 the investments in data centers, the new algorithms, the much larger models are 154 00:08:20,240 --> 00:08:22,717 entirely focused on that. So, we're not done. 155 00:08:22,800 --> 00:08:25,677 >> Okay. Um, Americans seem obsessed with 156 00:08:25,760 --> 00:08:28,717 China and China's developing AI at 157 00:08:28,800 --> 00:08:32,717 better pace and maybe in better ways than the United States. How do you today 158 00:08:32,800 --> 00:08:37,597 rate the competitive abilities of the Americans versus the Chinese on AI 159 00:08:37,680 --> 00:08:39,437 development? 160 00:08:39,520 --> 00:08:42,877 >> So, um, the US invented it and the 161 00:08:42,960 --> 00:08:45,117 Chinese have adopted it in a very clever 162 00:08:45,200 --> 00:08:48,957 way. So historically there was a view 163 00:08:49,040 --> 00:08:51,437 that China was well behind and China has 164 00:08:51,520 --> 00:08:54,157 in the last few years caught up really 165 00:08:54,240 --> 00:08:56,397 really well. Uh let me give you some 166 00:08:56,480 --> 00:08:59,117 examples. There are three models Kimmy 167 00:08:59,200 --> 00:09:02,477 Quinn and Deepseek which are often at 168 00:09:02,560 --> 00:09:07,357 the top of the rankings. This makes the Americans annoyed because Americans 169 00:09:07,440 --> 00:09:11,277 claim that the Chinese did something called distillation and used the 170 00:09:11,360 --> 00:09:14,877 American models to train them. That may or may not be true. The important thing 171 00:09:14,960 --> 00:09:17,357 is the Chinese models are really good. 172 00:09:17,440 --> 00:09:21,437 Furthermore, China has decided to be completely self-sufficient in this 173 00:09:21,520 --> 00:09:25,757 technology. So, they're building the entire hardware and software stack. They 174 00:09:25,840 --> 00:09:28,157 have their own chips called an SN930. 175 00:09:28,240 --> 00:09:32,397 They have their own software layers and so forth. The Chinese model is also 176 00:09:32,480 --> 00:09:36,877 different. They're not going for AGI or general intelligence. They're going for 177 00:09:36,960 --> 00:09:39,197 everything AI used everywhere in the 178 00:09:39,280 --> 00:09:41,757 society. And it's a massive massive 179 00:09:41,840 --> 00:09:44,157 program within the country to use AI 180 00:09:44,240 --> 00:09:47,037 everywhere especially on the edge. They 181 00:09:47,120 --> 00:09:52,077 are in my view six to 12 months behind which is not much. 182 00:09:52,160 --> 00:09:53,837 >> Now you've been an adviser to the 183 00:09:53,920 --> 00:09:56,477 defense department in a number of for a 184 00:09:56,560 --> 00:09:59,997 number of years. U how has AI changed 185 00:10:00,080 --> 00:10:03,517 warfare or is it going to change warfare? 186 00:10:03,600 --> 00:10:06,957 >> So the the the US government started a 187 00:10:07,040 --> 00:10:11,437 program called Maven. You'll you all will hear remember this about 10 years 188 00:10:11,520 --> 00:10:13,437 ago and I was involved with it because I 189 00:10:13,520 --> 00:10:15,997 was a consultant to the Pentagon during 190 00:10:16,080 --> 00:10:19,837 that period and Maven was can be 191 00:10:19,920 --> 00:10:24,077 understood as uh drones are running 192 00:10:24,160 --> 00:10:26,477 around looking at targets in the in in 193 00:10:26,560 --> 00:10:29,357 the at time in the military and those 194 00:10:29,440 --> 00:10:32,077 video feeds were being watched. So the 195 00:10:32,160 --> 00:10:34,077 first use in the military is actually 196 00:10:34,160 --> 00:10:38,717 just surveillance. Why? Most soldiers in 197 00:10:38,800 --> 00:10:40,557 the military are just watching for 198 00:10:40,640 --> 00:10:43,677 things. That was the first and best use. 199 00:10:43,760 --> 00:10:47,117 The next one will be drones. And because 200 00:10:47,200 --> 00:10:48,957 of what's going on in Ukraine, the US 201 00:10:49,040 --> 00:10:51,997 government has decided to to learn from 202 00:10:52,080 --> 00:10:56,477 Ukraine. Now there's a Secretary Hexith has announced a big program around 203 00:10:56,560 --> 00:10:58,957 drones and drone adoption, land, sea, 204 00:10:59,040 --> 00:11:02,237 and air. Okay. Um, if you were advising 205 00:11:02,320 --> 00:11:06,877 a young person today who's graduating from college, uh, should they join a 206 00:11:06,960 --> 00:11:11,037 large accounting consulting firm like yours and have an accounting background 207 00:11:11,120 --> 00:11:15,357 as you do or should they get a technology background in AI? Which one 208 00:11:15,440 --> 00:11:19,757 is going to be more exciting and make more money for that young person? 209 00:11:19,840 --> 00:11:23,517 I actually don't think this is an eitheror and that in fact what we're 210 00:11:23,600 --> 00:11:27,277 finding in the application space I agree with you Eric that we're just at the 211 00:11:27,360 --> 00:11:31,677 beginning of the wave of AI adoption and I'll use my baseball analogy we're in 212 00:11:31,760 --> 00:11:33,837 the the top of the first inning here 213 00:11:33,920 --> 00:11:36,477 those who have the ability to integrate 214 00:11:36,560 --> 00:11:38,717 across disciplines. So to really bring 215 00:11:38,800 --> 00:11:41,437 this to life in an enterprise, you have 216 00:11:41,520 --> 00:11:46,157 to both have some baseline understanding of the technology, the capability of the 217 00:11:46,240 --> 00:11:50,557 model, how to prompt it to yield the desired outcome, as well as the business 218 00:11:50,640 --> 00:11:54,717 problem you're trying to solve if you're trying to embed this into the finance 219 00:11:54,800 --> 00:11:56,957 function, the manufacturing function, 220 00:11:57,040 --> 00:11:59,277 customer and marketing. And where we've 221 00:11:59,360 --> 00:12:02,477 seen things go wrong is where you take a 222 00:12:02,560 --> 00:12:05,117 piece of technology that's developed by 223 00:12:05,200 --> 00:12:08,077 someone in the tech laboratory and you 224 00:12:08,160 --> 00:12:11,197 try to deploy that and force it into an 225 00:12:11,280 --> 00:12:16,157 enterprise. Generally speaking, it fails. You need the people in the heart 226 00:12:16,240 --> 00:12:18,077 of the business who understand that 227 00:12:18,160 --> 00:12:19,997 function, who understand that industry 228 00:12:20,080 --> 00:12:25,437 and sector to be paired up on a crossf functional team to design something 229 00:12:25,520 --> 00:12:28,717 that's fit for purpose in the real world. I thought Yan's words earlier 230 00:12:28,800 --> 00:12:31,517 were were very accurate. The real world 231 00:12:31,600 --> 00:12:35,997 is far more complicated than text. And so those who have an appreciation for 232 00:12:36,080 --> 00:12:40,717 the context and who can work with the technologists to bring it to life, 233 00:12:40,800 --> 00:12:43,117 that's where the real value is being created. 234 00:12:43,200 --> 00:12:44,797 >> Let's suppose somebody in the audience 235 00:12:44,880 --> 00:12:48,477 here is uh 40 years old. They majored in 236 00:12:48,560 --> 00:12:51,197 humanities in college. they have a job 237 00:12:51,280 --> 00:12:56,157 that's uh say a non- tech job, but they really want to learn about AI so they 238 00:12:56,240 --> 00:12:59,917 can talk about it intelligently with their children, if not somebody else. 239 00:13:00,000 --> 00:13:04,557 What's the best way for a middle-aged non-technology person to actually learn 240 00:13:04,640 --> 00:13:06,717 something about AI to sound reasonably 241 00:13:06,800 --> 00:13:09,117 intelligent on it? 242 00:13:09,200 --> 00:13:12,237 >> So, let me turn the question around. We 243 00:13:12,320 --> 00:13:14,797 have MIT and Stanford here. The most 244 00:13:14,880 --> 00:13:18,637 important new course that you don't have 245 00:13:18,720 --> 00:13:22,237 is the freshman class on prompt engineering. 246 00:13:22,320 --> 00:13:25,837 Because if you think about it, the student that is entering MIT or 247 00:13:25,920 --> 00:13:27,517 Stanford, 248 00:13:27,600 --> 00:13:30,637 their entire life is going to be prompt 249 00:13:30,720 --> 00:13:34,877 engineering as it evolves because everything they do will be with a 250 00:13:34,960 --> 00:13:37,437 computer assisting them. So the same 251 00:13:37,520 --> 00:13:39,837 applies to the 40-year-old, right? So 252 00:13:39,920 --> 00:13:42,237 the 40-year-old needs to get good, too. 253 00:13:42,320 --> 00:13:44,797 These tools are remarkably powerful for 254 00:13:44,880 --> 00:13:47,357 a non-technical person. Um, if you 255 00:13:47,440 --> 00:13:49,437 wanted to write code, you wanted to do 256 00:13:49,520 --> 00:13:52,557 something, have fun, learn how to do it. 257 00:13:52,640 --> 00:13:55,197 They will produce working systems that 258 00:13:55,280 --> 00:13:58,237 can make your life better. Okay. So 259 00:13:58,320 --> 00:14:02,797 today um a lot of the big technology companies that are uh have trillion 260 00:14:02,880 --> 00:14:04,477 dollar market caps, they seem to be 261 00:14:04,560 --> 00:14:06,397 moving away from everybody else in terms 262 00:14:06,480 --> 00:14:10,077 of their ability to um develop AI models 263 00:14:10,160 --> 00:14:13,917 and the ability be a leader in it. What about the smaller companies? Is there a 264 00:14:14,000 --> 00:14:15,917 chance for a smaller company today to 265 00:14:16,000 --> 00:14:19,037 become uh a gigantic uh alphabet or is 266 00:14:19,120 --> 00:14:22,317 that day gone already too late to become 267 00:14:22,400 --> 00:14:24,637 gigantic again in the technology world? 268 00:14:24,720 --> 00:14:28,477 So, um, the big companies are getting 269 00:14:28,560 --> 00:14:32,397 bigger because their revenue is growing fast. And it's growing fast because 270 00:14:32,480 --> 00:14:35,277 they've been around for a while and they have a lot of capital and they have a 271 00:14:35,360 --> 00:14:38,957 lot of data centers, right? That's what's really going on. They're not just 272 00:14:39,040 --> 00:14:40,637 being nice and the valuations are going 273 00:14:40,720 --> 00:14:44,157 up. So, for smaller companies, the real 274 00:14:44,240 --> 00:14:48,397 question is how do they get the capital to play in that game? It's going to get 275 00:14:48,480 --> 00:14:50,397 harder and harder and harder as the 276 00:14:50,480 --> 00:14:52,957 industry matures. And let me ask you 277 00:14:53,040 --> 00:14:55,837 about Nvidia. And Nvidia is making a lot 278 00:14:55,920 --> 00:15:00,157 of money on artificial intelligence. Is anybody else really making money or are 279 00:15:00,240 --> 00:15:03,597 they just spending money on artificial intelligence? 280 00:15:03,680 --> 00:15:05,437 Look, when you when you look at where 281 00:15:05,520 --> 00:15:09,037 the rents have accred to date, obviously 282 00:15:09,120 --> 00:15:13,357 the sort of infrastructure layer at the front end as people are building all of 283 00:15:13,440 --> 00:15:15,197 this out has captured a a 284 00:15:15,280 --> 00:15:18,237 disproportionate share. The real issue 285 00:15:18,320 --> 00:15:20,797 from a societal standpoint is how to 286 00:15:20,880 --> 00:15:23,597 take those gains and spread them broadly 287 00:15:23,680 --> 00:15:25,837 and actually drive the productivity 288 00:15:25,920 --> 00:15:28,797 benefit into the heart of the of the 289 00:15:28,880 --> 00:15:32,317 enterprise system. And frankly, it's a 290 00:15:32,400 --> 00:15:34,637 societal imperative if we're going to 291 00:15:34,720 --> 00:15:37,757 earn the confidence of people that in an 292 00:15:37,840 --> 00:15:41,917 environment where so much fear is being attached to the technology and some of 293 00:15:42,000 --> 00:15:44,237 the risks, the sort of misinformation, 294 00:15:44,320 --> 00:15:46,877 the potential job displacement, and that 295 00:15:46,960 --> 00:15:49,437 just spurs up backlash. It spurs up 296 00:15:49,520 --> 00:15:51,997 political challenge. We have to be able 297 00:15:52,080 --> 00:15:57,517 to demonstrate that for the average citizen, benefits are going to acrue, 298 00:15:57,600 --> 00:16:00,477 that there's going to be incremental economic opportunities, that there's 299 00:16:00,560 --> 00:16:03,757 going to be improvements to the quality of human life, that there's going to be 300 00:16:03,840 --> 00:16:05,757 some incremental value in terms of 301 00:16:05,840 --> 00:16:10,797 educating their children. And so that process of taking the technology and 302 00:16:10,880 --> 00:16:14,317 making it work in the lives of real people, we're in the really early 303 00:16:14,400 --> 00:16:16,077 stages. All >> right. Let me ask you, for your company, 304 00:16:16,160 --> 00:16:20,477 you have 500,000 employees roughly. All right. Suppose somebody wants to be an 305 00:16:20,560 --> 00:16:24,397 employee in your company now. They're graduating from college and they are in 306 00:16:24,480 --> 00:16:29,117 a literature major or they majored in social science subjects or humanities 307 00:16:29,200 --> 00:16:33,197 subjects. They have any chance of getting hired. And if you get hired by 308 00:16:33,280 --> 00:16:36,877 your firm, do you have to be a STEM major today? And how do these non- STEM 309 00:16:36,960 --> 00:16:40,397 majors fair? Do they actually wind up staying there after a year or two or 310 00:16:40,480 --> 00:16:43,277 they get washed out? I want best 311 00:16:43,360 --> 00:16:45,757 available athletes. Now they have to 312 00:16:45,840 --> 00:16:48,637 want to acquire the skill sets that Eric 313 00:16:48,720 --> 00:16:50,717 shared in terms of the prompt 314 00:16:50,800 --> 00:16:55,677 engineering in terms of understanding the core of the technology and how that 315 00:16:55,760 --> 00:16:59,357 sort of applies in terms of a modern business enterprise. But those who were 316 00:16:59,440 --> 00:17:04,317 deep thinkers and in fact those who can best understand the societal 317 00:17:04,400 --> 00:17:07,357 implications of this and how you actually adapt the use of the technology 318 00:17:07,440 --> 00:17:11,117 in an ethical fashion. I'm still a believer that the generalist has a big 319 00:17:11,200 --> 00:17:11,837 role here. 320 00:17:11,920 --> 00:17:14,877 >> Okay. Eric, you were the CEO of Noville, 321 00:17:14,960 --> 00:17:18,157 right? And then uh two young guys were 322 00:17:18,240 --> 00:17:22,797 looking for a CEO for a new company they were starting that came the name Google. 323 00:17:22,880 --> 00:17:24,317 Uh first of all, why did they spell 324 00:17:24,400 --> 00:17:26,557 Google improperly? 325 00:17:26,640 --> 00:17:30,397 So u there's a number called Google G O 326 00:17:30,480 --> 00:17:34,637 G O L 10 to the 100th. And the trademark 327 00:17:34,720 --> 00:17:39,597 was already taken. So Sergey changed it to LE and the rest is history. We 328 00:17:39,680 --> 00:17:43,277 ultimately settled with the family of the mathematician who invented the 329 00:17:43,360 --> 00:17:45,517 Google um because they were slightly 330 00:17:45,600 --> 00:17:48,077 annoyed at our reframing of that number. 331 00:17:48,160 --> 00:17:52,477 >> When you joined the company um people said there are a lot of search engines 332 00:17:52,560 --> 00:17:57,277 out here. Why do we need one more search engine? What was Google's ability to do 333 00:17:57,360 --> 00:18:00,637 something in search that nobody else had done? Was it a better system or better 334 00:18:00,720 --> 00:18:03,837 mousetrap? Um, so if you look at 335 00:18:03,920 --> 00:18:07,517 technology these days, every wave has 336 00:18:07,600 --> 00:18:10,717 five or 10 companies that are similar. 337 00:18:10,800 --> 00:18:13,437 In Google's case, it was the last of a 338 00:18:13,520 --> 00:18:15,837 series of companies that were doing 339 00:18:15,920 --> 00:18:20,317 ranking. But Larry Paige had invented in 340 00:18:20,400 --> 00:18:22,957 his dorm room an algorithm called page 341 00:18:23,040 --> 00:18:27,597 rank which is mathematically a fastforier transform of the importance 342 00:18:27,680 --> 00:18:32,717 of something and he used it for ranking and Google's ranking was just better. He 343 00:18:32,800 --> 00:18:37,197 started it in the dor dorm room in Stanford on the campus started to use it. 344 00:18:37,280 --> 00:18:40,557 >> Now one of the things you had at Google at the very beginning was something when 345 00:18:40,640 --> 00:18:43,037 I toured it you gave me a tour early on. 346 00:18:43,120 --> 00:18:47,197 You had free food. Um, and that was a I 347 00:18:47,280 --> 00:18:50,557 guess a way to keep people on the campus and not waste time going out for 348 00:18:50,640 --> 00:18:53,597 restaurants. Was that a good idea? Whose idea was that? 349 00:18:53,680 --> 00:18:56,077 >> So, so Sergey and Larry, and I inherited 350 00:18:56,160 --> 00:18:58,797 this, believe that a company was like a 351 00:18:58,880 --> 00:19:01,677 family and that you should feed them. 352 00:19:01,760 --> 00:19:04,237 Um, and you should feed them together because that's how they would all 353 00:19:04,320 --> 00:19:08,317 communicate. They had a number of strange beliefs. for example, they loved 354 00:19:08,400 --> 00:19:13,837 hiring Olympians because they figured that Olympians were so disciplined to to 355 00:19:13,920 --> 00:19:17,357 become an Olympian that they would be great employees even if they knew 356 00:19:17,440 --> 00:19:20,477 nothing about the job that they were working on. That turned out to be true, 357 00:19:20,560 --> 00:19:25,837 too. Um, so so in any case, the the free food has continued and I think it's now 358 00:19:25,920 --> 00:19:29,437 standard in >> they also had a thing called 20% time. 359 00:19:29,520 --> 00:19:31,757 You could spend 20% of your time doing 360 00:19:31,840 --> 00:19:33,837 something else. Would that work out? 361 00:19:33,920 --> 00:19:36,717 >> Uh, it did. Um, and so Google had a rule 362 00:19:36,800 --> 00:19:41,437 that every employee, at least the technical ones, could spend 20% of their 363 00:19:41,520 --> 00:19:45,357 time on what they wanted and not what their manager was. Now, you sit there 364 00:19:45,440 --> 00:19:47,277 and you go, how is that make any sense 365 00:19:47,360 --> 00:19:51,597 at all? But the way it worked was if the 366 00:19:51,680 --> 00:19:54,957 boss came in and said you have to spend 367 00:19:55,040 --> 00:19:57,357 all of your time on this and was very 368 00:19:57,440 --> 00:20:00,397 manipulative. It was a control on the 369 00:20:00,480 --> 00:20:02,317 power of the management because you 370 00:20:02,400 --> 00:20:04,957 could always say yes I'll give you four 371 00:20:05,040 --> 00:20:09,997 fifths of my day of my year my week but I'm not going to give you all of it. It 372 00:20:10,080 --> 00:20:13,837 was a balance of power between the creative engineers and their management. 373 00:20:13,920 --> 00:20:15,357 And a final question about the early 374 00:20:15,440 --> 00:20:18,637 days of this company. You were the CEO 375 00:20:18,720 --> 00:20:23,277 when you went on a trip. Somebody moved into your office. You're the CEO and 376 00:20:23,360 --> 00:20:27,037 somebody took your office and you couldn't get him to leave. Why was that 377 00:20:27,120 --> 00:20:30,317 a good thing to do? >> Well, I I had left I had left town for 378 00:20:30,400 --> 00:20:34,797 the for a sales call and I came back and I found a fellow named Emit in my 379 00:20:34,880 --> 00:20:39,597 office. He was at 8 by12 office sitting next to my little table. And I said, 380 00:20:39,680 --> 00:20:43,677 "Who are you?" And he identified himself and I said, "Um, who gave you 381 00:20:43,760 --> 00:20:46,957 permission?" And I said, "I asked Wayne who worked for me." and Wayne said, 382 00:20:47,040 --> 00:20:49,517 "Sure, move into Eric's office." Now, 383 00:20:49,600 --> 00:20:52,557 this is a test as an executive because 384 00:20:52,640 --> 00:20:55,197 when this happens, you can say, "Get the 385 00:20:55,280 --> 00:20:57,277 hell out of my office." Which is 386 00:20:57,360 --> 00:21:00,317 culturally not correct. So, I figured, 387 00:21:00,400 --> 00:21:05,037 okay, I'll just sit there and see what happens. So, he wore he was a programmer 388 00:21:05,120 --> 00:21:09,037 and he wore headphones. And I used to talk on the phone and I was talking to 389 00:21:09,120 --> 00:21:12,797 the sales executive at the time. We didn't know what our revenue was. And he 390 00:21:12,880 --> 00:21:16,877 said, "I think the revenue is going to be 115 million for the quarter. Imagine 391 00:21:16,960 --> 00:21:19,037 these numbers." And I said, "I think 392 00:21:19,120 --> 00:21:21,117 you're sandbagging," which is a term in 393 00:21:21,200 --> 00:21:24,717 sales for not telling the truth. Um, so 394 00:21:24,800 --> 00:21:28,477 I have the conversation. I don't make very much progress. And so I put the 395 00:21:28,560 --> 00:21:32,237 phone down and a meet takes his headphones off and says, "I can tell you 396 00:21:32,320 --> 00:21:36,157 the revenue number." And I said, "I knew you were listening on my phone." He'd 397 00:21:36,240 --> 00:21:38,637 been listening to all of my phone calls. 398 00:21:38,720 --> 00:21:40,397 And so what happens is I said, 'What's 399 00:21:40,480 --> 00:21:44,397 the revenue number? And he said 140. And 400 00:21:44,480 --> 00:21:48,157 so I made Omid present all of his sales 401 00:21:48,240 --> 00:21:52,397 forecastes standing on a sand sandbags 402 00:21:52,480 --> 00:21:55,117 as punishment for lying to me. 403 00:21:55,200 --> 00:21:57,437 >> So has anybody ever squatted in your office? 404 00:21:57,520 --> 00:22:01,277 >> Uh not yet. I'm old. >> My advice to those of you as CEO is get 405 00:22:01,360 --> 00:22:06,557 the person doing the revenue in the company in your office and watching it. 406 00:22:06,640 --> 00:22:10,317 And the simplest rule for those of you who are CEOs is the only thing that 407 00:22:10,400 --> 00:22:13,197 matters in the CEO's life is revenue. 408 00:22:13,280 --> 00:22:17,917 Solves all problems. Now you're running I'm sorry. We became best friends. 409 00:22:18,000 --> 00:22:21,917 You're running Deote. How many people come into you and tell you bad news as 410 00:22:22,000 --> 00:22:23,437 opposed to people coming in to tell you 411 00:22:23,520 --> 00:22:25,677 good news? 412 00:22:25,760 --> 00:22:30,237 The ratio is undoubtedly higher good to 413 00:22:30,320 --> 00:22:34,877 bad, but not [clears throat] not as detrimental as you'd think. And in fact, 414 00:22:34,960 --> 00:22:37,117 I I actually think the number one 415 00:22:37,200 --> 00:22:40,077 dimension to effective leadership in an 416 00:22:40,160 --> 00:22:44,637 environment like we have today where there's so much disruption and so much 417 00:22:44,720 --> 00:22:47,037 uncertainty is creating a culture where 418 00:22:47,120 --> 00:22:51,757 there is a willingness to surface problems where you see a top- down 419 00:22:51,840 --> 00:22:53,437 command and control and there's a fear 420 00:22:53,520 --> 00:22:56,557 and all the dashboards are green or or 421 00:22:56,640 --> 00:22:58,877 that is a recipe for an organization 422 00:22:58,960 --> 00:23:01,437 that will be disrupted where you create 423 00:23:01,520 --> 00:23:03,917 a culture where people are actually 424 00:23:04,000 --> 00:23:08,957 willing to keep an open mind even if the strategy said X, the market is shifting, 425 00:23:09,040 --> 00:23:11,917 the environment is shifting, the technology is shifting, we're going to 426 00:23:12,000 --> 00:23:17,037 challenge our assumption, we're going to be sort of embodying a learning mindset. 427 00:23:17,120 --> 00:23:19,757 Those are the organizations that are actually going to thrive today. So you 428 00:23:19,840 --> 00:23:23,597 you need a lot more willingness to hear things that are unpleasant. And let me 429 00:23:23,680 --> 00:23:25,197 ask you about artificial intelligence 430 00:23:25,280 --> 00:23:27,917 this respect today in colleges when 431 00:23:28,000 --> 00:23:30,717 people get exams uh that say they're 432 00:23:30,800 --> 00:23:33,197 take-home exams uh and they write them 433 00:23:33,280 --> 00:23:35,837 up. How do they professors know that 434 00:23:35,920 --> 00:23:38,397 they're really not written by AI? And as 435 00:23:38,480 --> 00:23:43,117 I understand it, there is now some software that can determine whether it's 436 00:23:43,200 --> 00:23:47,037 been written by AI, but there's software that can determine that the how to fool 437 00:23:47,120 --> 00:23:51,037 the professor's software. How does that work? I've got two kids in college right 438 00:23:51,120 --> 00:23:53,917 now and this dominated the conversation 439 00:23:54,000 --> 00:23:57,037 over the winter break. The worst thing 440 00:23:57,120 --> 00:23:58,957 that an educational institution could 441 00:23:59,040 --> 00:24:03,117 possibly do is ban the usage of AI, ban 442 00:24:03,200 --> 00:24:07,837 the usage of the tool that's actually going to help make the individual more 443 00:24:07,920 --> 00:24:11,997 effective. And yet you have this huge dilemma because if all the student is 444 00:24:12,080 --> 00:24:16,317 doing is putting the assignment into the AI engine and turning in something that 445 00:24:16,400 --> 00:24:18,717 the LLM spit out, that is a useless 446 00:24:18,800 --> 00:24:20,717 earning learning exercise. And so how 447 00:24:20,800 --> 00:24:24,557 you actually allow for the usage but 448 00:24:24,640 --> 00:24:27,517 still demand the human overlay and the 449 00:24:27,600 --> 00:24:31,757 critical thinking and how did I take the output and use it to push myself. And 450 00:24:31,840 --> 00:24:34,157 frankly, these software programs that 451 00:24:34,240 --> 00:24:36,717 are detecting what percentage was 452 00:24:36,800 --> 00:24:41,757 actually derived by AI, they have some directional sense, but there's a lot of 453 00:24:41,840 --> 00:24:45,597 false positives. >> How has your job changed by AI? Do you 454 00:24:45,680 --> 00:24:48,237 use AI yourself? You go and chat GBT all 455 00:24:48,320 --> 00:24:50,637 the time to get answers to things. 456 00:24:50,720 --> 00:24:54,077 >> Look, I I use it constantly in my 457 00:24:54,160 --> 00:24:57,997 personal life. I find it actually makes me much more effective at any variety of 458 00:24:58,080 --> 00:24:59,197 things that we come across in our 459 00:24:59,280 --> 00:25:03,357 day-to-day life. And I am embracing and 460 00:25:03,440 --> 00:25:07,357 trying to lead by example in terms of the tools, the demos, the proofs of 461 00:25:07,440 --> 00:25:10,077 concept, and then ultimately those things that look like they're going to 462 00:25:10,160 --> 00:25:15,357 scale that we're putting into production of using them myself, of providing 463 00:25:15,440 --> 00:25:18,877 feedback in terms of trying to help train the model so that where there are 464 00:25:18,960 --> 00:25:23,197 things that are valuable, feed that back and let's capitalize it. And also 465 00:25:23,280 --> 00:25:27,517 calling out the meter that things that look great on a screen and 466 00:25:27,600 --> 00:25:31,277 to an average uh recipient of the information appear to be very 467 00:25:31,360 --> 00:25:34,797 impressive, but to somebody who's an expert in the area is a bunch of 468 00:25:34,880 --> 00:25:39,037 gibberish. And that's the real dilemma right now. This stuff looks really 469 00:25:39,120 --> 00:25:43,837 impressive on the surface. And some of it is in substance and some of it's 470 00:25:43,920 --> 00:25:45,757 garbage. And it takes the human who has 471 00:25:45,840 --> 00:25:47,917 the domain knowledge to actually pick 472 00:25:48,000 --> 00:25:51,037 between the two. Eric, uh, bad news 473 00:25:51,120 --> 00:25:53,357 usually sells better than good news. 474 00:25:53,440 --> 00:25:57,917 People like to read bad news, it seems, and that's why CNN or the equivalent 475 00:25:58,000 --> 00:26:00,077 shows they when they have a crisis, it 476 00:26:00,160 --> 00:26:02,797 ratings go up and so forth. Um can you 477 00:26:02,880 --> 00:26:06,237 to lead finish this uh conversation by 478 00:26:06,320 --> 00:26:10,477 making people feel that AI is going to be a good thing for civilization not a 479 00:26:10,560 --> 00:26:13,037 bad thing and let people leave here 480 00:26:13,120 --> 00:26:17,517 thinking that if they know more about AI they'll be better off and living in a 481 00:26:17,600 --> 00:26:19,517 world with AI being more and more 482 00:26:19,600 --> 00:26:21,757 important is a good thing for everybody. 483 00:26:21,840 --> 00:26:26,397 >> So we have some choices as humans and 484 00:26:26,480 --> 00:26:28,237 those choices are really important to 485 00:26:28,320 --> 00:26:33,597 get right. So first place, this technology is going to happen. It's not 486 00:26:33,680 --> 00:26:36,717 going to get prevented. It's not going to get stopped. There's too many 487 00:26:36,800 --> 00:26:38,877 countries, too many people, too many 488 00:26:38,960 --> 00:26:41,277 incentives. It's going to happen. So 489 00:26:41,360 --> 00:26:43,997 what does this mean? It means great 490 00:26:44,080 --> 00:26:46,717 great solutions for healthcare, new 491 00:26:46,800 --> 00:26:49,437 drugs, better energy solutions, better 492 00:26:49,520 --> 00:26:51,597 power distribution, better physics, 493 00:26:51,680 --> 00:26:53,757 better chemistry, better sociology, 494 00:26:53,840 --> 00:26:57,037 better research. It also means that it 495 00:26:57,120 --> 00:27:01,277 can be used for bad. It can be used for example for oppression. It can be used 496 00:27:01,360 --> 00:27:03,437 to limit freedom in governments 497 00:27:03,520 --> 00:27:05,677 hopefully not in the west. Uh it can be 498 00:27:05,760 --> 00:27:08,957 used in war. Um the technology itself is 499 00:27:09,040 --> 00:27:12,877 so addictive that it can affect our young people and we should make sure 500 00:27:12,960 --> 00:27:15,837 that our young people are protected from some of the worst parts of the 501 00:27:15,920 --> 00:27:18,717 technology. I could go on. But we face 502 00:27:18,800 --> 00:27:20,797 choices. I've written two books on this. 503 00:27:20,880 --> 00:27:23,997 We face choices now about how we want to 504 00:27:24,080 --> 00:27:27,837 deal with this incredibly powerful technology. I will tell you and it's 505 00:27:27,920 --> 00:27:31,917 really important to understand that we are living through a moment that will be 506 00:27:32,000 --> 00:27:34,557 in history for thousands of years. It is 507 00:27:34,640 --> 00:27:37,757 the moment in history when a nonhuman 508 00:27:37,840 --> 00:27:42,797 intelligence arrived and it was a competitor to us. Right? How we use 509 00:27:42,880 --> 00:27:46,477 that, how we shape it, right? How we guarantee alignment will determine the 510 00:27:46,560 --> 00:27:48,797 outcome of your >> final question. If you had the chance to 511 00:27:48,880 --> 00:27:53,277 rename artificial intelligence and give it a name that's more appropriate, would 512 00:27:53,360 --> 00:27:56,317 you change the name or would you just say it's better? The best name is 513 00:27:56,400 --> 00:27:59,277 artificial intelligence. >> The the best name is artificial 514 00:27:59,360 --> 00:28:03,837 intelligence because it is intelligence and it's artificial. Remember that its 515 00:28:03,920 --> 00:28:08,237 intelligence is not the same as human intelligence. It doesn't have the same 516 00:28:08,320 --> 00:28:10,397 values. It doesn't think the same way. 517 00:28:10,480 --> 00:28:12,717 It doesn't think at all. It's a it's a 518 00:28:12,800 --> 00:28:15,517 different kind of intelligence. When we 519 00:28:15,600 --> 00:28:18,077 work with it, we will be more powerful. 520 00:28:18,160 --> 00:28:21,037 When it runs us, we will be very unhappy. 521 00:28:21,120 --> 00:28:25,597 >> Is it harder to learn artificial intelligence or pass the CPA exam? 522 00:28:25,680 --> 00:28:28,317 Well, those two are arguably becoming one and the same. That's one of the 523 00:28:28,400 --> 00:28:30,157 tests that's been used to define the 524 00:28:30,240 --> 00:28:32,637 efficacy of the technology. But we do 525 00:28:32,720 --> 00:28:37,037 have a real issue of education around 526 00:28:37,120 --> 00:28:41,037 the societal benefit. And there's some recent polling where those who are 527 00:28:41,120 --> 00:28:43,437 actually using the technology themselves 528 00:28:43,520 --> 00:28:46,557 are far more positive on the potential 529 00:28:46,640 --> 00:28:49,037 impact than those who aren't using it. 530 00:28:49,120 --> 00:28:51,357 And I I do worry, Eric, the technology 531 00:28:51,440 --> 00:28:54,157 is coming. it is going to happen. But 532 00:28:54,240 --> 00:28:56,237 we're in this powder keg of a moment 533 00:28:56,320 --> 00:28:58,797 society where you see the sort of the 534 00:28:58,880 --> 00:29:01,197 populist concerns and the fears over the 535 00:29:01,280 --> 00:29:03,757 future and then you layer this onto it 536 00:29:03,840 --> 00:29:06,397 and all the worries over job disruption. 537 00:29:06,480 --> 00:29:10,477 Our number one job as leaders is to actually go out and tell the story and 538 00:29:10,560 --> 00:29:14,637 make this real in terms of how this is going to improve the quality of human 539 00:29:14,720 --> 00:29:17,437 life. >> Thank you both for a very interesting 540 00:29:17,520 --> 00:29:20,797 conversation. How many people here feel they know a lot more about AI than they 541 00:29:20,880 --> 00:29:23,997 did a half hour ago? >> Okay, thank you very much. 542 00:29:24,080 --> 00:29:27,800 >> Thank you, Dave. Thank you.