1 00:00:04,580 --> 00:00:06,600 [music] 2 00:00:08,320 --> 00:00:11,320 Okay. 3 00:00:13,520 --> 00:00:16,477 >> Hey everyone. 4 00:00:16,560 --> 00:00:19,437 >> All right. Um, so I'm pleased to 5 00:00:19,520 --> 00:00:23,757 introduce Elon Musk today. Um, he doesn't need I think an introduction, 6 00:00:23,840 --> 00:00:25,917 but I just remind you all of the 7 00:00:26,000 --> 00:00:29,517 businesses he's involved in. Tesla 8 00:00:29,600 --> 00:00:31,917 that's reinventing the automotive 9 00:00:32,000 --> 00:00:35,517 industry. SpaceX sending 80% of the 10 00:00:35,600 --> 00:00:38,717 world's payload into orbit. Starlink, 11 00:00:38,800 --> 00:00:41,677 Neurolink, the Boring Company. I could 12 00:00:41,760 --> 00:00:44,397 go on. He's also, we know, the owner of 13 00:00:44,480 --> 00:00:49,277 X, uh, company formerly known as Twitter. 14 00:00:49,360 --> 00:00:52,317 >> Never heard of it. >> You heard of it? 15 00:00:52,400 --> 00:00:54,157 Statistically, 16 00:00:54,240 --> 00:00:57,837 uh, 560 monthly average users, 34 17 00:00:57,920 --> 00:01:01,757 minutes a day, 75% under 35. So, a 18 00:01:01,840 --> 00:01:05,757 pretty important platform in our industry. And, you know, I don't agree 19 00:01:05,840 --> 00:01:07,757 with Elon on everything. >> What? 20 00:01:07,840 --> 00:01:11,917 >> Um, no. >> Come on. >> But he owns an important platform and I 21 00:01:12,000 --> 00:01:14,717 thought we should listen to him and understand a little bit about what he's 22 00:01:14,800 --> 00:01:17,677 trying to do. So, we're going to talk this morning about X. Hopefully we'll 23 00:01:17,760 --> 00:01:22,157 talk a bit about AI and then about innovation and creativity and why you 24 00:01:22,240 --> 00:01:26,157 get up in the morning. >> What what what do you disagree with me? 25 00:01:26,240 --> 00:01:28,397 >> Creativity. [laughter] 26 00:01:28,480 --> 00:01:30,397 AI is not going to destroy that. 27 00:01:30,480 --> 00:01:33,197 >> Um so Elon, welcome to the heart of the 28 00:01:33,280 --> 00:01:36,957 advertising industry. Now back in um 29 00:01:37,040 --> 00:01:39,917 November, you had a you had a message to 30 00:01:40,000 --> 00:01:46,600 us. Um you told us to sort of go ourselves 31 00:01:47,200 --> 00:01:52,077 So maybe we start there. 32 00:01:52,160 --> 00:01:56,637 >> Why? It's a serious question. It's a It's a serious question. Why did you say 33 00:01:56,720 --> 00:01:57,517 that? 34 00:01:57,600 --> 00:02:01,597 >> And what did you mean by that? 35 00:02:01,680 --> 00:02:06,920 Well, um 36 00:02:08,080 --> 00:02:12,797 well, first of all, it wasn't a um you know, it wasn't to the 37 00:02:12,880 --> 00:02:14,877 advertising to ad advertisers as a 38 00:02:14,960 --> 00:02:18,877 whole. It was uh with respect to freedom 39 00:02:18,960 --> 00:02:21,357 of speech. I think there it is important 40 00:02:21,440 --> 00:02:23,437 to have a a global free speech platform 41 00:02:23,520 --> 00:02:26,397 where people from a wide range of 42 00:02:26,480 --> 00:02:31,117 opinions can voice their their views. Um 43 00:02:31,200 --> 00:02:34,157 and um and in some cases uh there were 44 00:02:34,240 --> 00:02:39,677 advertisers who weren't uh insisting on censorship and at and at 45 00:02:39,760 --> 00:02:42,557 the end of the day if there is censorship 46 00:02:42,640 --> 00:02:44,077 uh if we have to make a choice between 47 00:02:44,160 --> 00:02:47,517 censorship and money you know censorship 48 00:02:47,600 --> 00:02:52,077 and money or free speech and losing money. We're going to pick the second. 49 00:02:52,160 --> 00:02:53,277 we're gonna we're going to support free 50 00:02:53,360 --> 00:02:57,357 speech uh rather than than agree to be 51 00:02:57,440 --> 00:03:00,957 censored for for money, which is um I 52 00:03:01,040 --> 00:03:03,517 think the right moral decision. Now, of 53 00:03:03,600 --> 00:03:06,717 course, uh advertisers have a right to 54 00:03:06,800 --> 00:03:10,317 appear next to content that they find uh 55 00:03:10,400 --> 00:03:14,397 compatible with their brands. That's totally fine. I think that that is a 56 00:03:14,480 --> 00:03:18,877 that that's that's again a choice of an advertiser to appear next to content 57 00:03:18,960 --> 00:03:21,117 that they think fits with their brand. 58 00:03:21,200 --> 00:03:25,757 That's totally cool. But what what is not cool is insisting that that there 59 00:03:25,840 --> 00:03:28,637 can be no content that they disagree with on the platform. 60 00:03:28,720 --> 00:03:32,717 >> So like other media proprietors who want to run the editorial policy of the 61 00:03:32,800 --> 00:03:37,517 company, you feel you know >> Yeah. I mean >> you shouldn't be under pressure to make 62 00:03:37,600 --> 00:03:41,117 that compromise between money and freedom of speech. 63 00:03:41,200 --> 00:03:43,757 >> Yes. It's I think there's an it's an 64 00:03:43,840 --> 00:03:48,637 important distinction here that uh 65 00:03:48,720 --> 00:03:52,237 we we don't want to have we we don't want to take money to censor. 66 00:03:52,320 --> 00:03:56,237 >> That would be to censor broadly in the platform. I think that would be wrong. 67 00:03:56,320 --> 00:03:58,717 Um I think free speech is the bedrock of 68 00:03:58,800 --> 00:04:03,597 of democracy. Um and if the people in 69 00:04:03,680 --> 00:04:06,477 order for X to be the the public square 70 00:04:06,560 --> 00:04:10,877 of the world, it's got to it's really got to be a free speech platform. Um now 71 00:04:10,960 --> 00:04:12,477 that doesn't mean people can say illegal 72 00:04:12,560 --> 00:04:14,877 things is free speech within the bounds 73 00:04:14,960 --> 00:04:17,917 of the law. Um but not not going further 74 00:04:18,000 --> 00:04:19,997 than the law uh for a particular 75 00:04:20,080 --> 00:04:23,677 country. Um I think it's imperative that 76 00:04:23,760 --> 00:04:27,437 the the people have a right to speak. Um 77 00:04:27,520 --> 00:04:32,637 uh but by the same token the adver you know advertisers have a 78 00:04:32,720 --> 00:04:36,637 right to appear next to content that they feel is compatible with their brand. 79 00:04:36,720 --> 00:04:39,277 >> Yeah. And that's the brand safety question which I going to ask you next 80 00:04:39,360 --> 00:04:43,597 which is you know to be clear you do you do understand and believe in brand 81 00:04:43,680 --> 00:04:48,797 safety in that sense. >> Yes. I mean I I I believe in 82 00:04:48,880 --> 00:04:53,757 sort of freedom of choice in the sense that advertisers have a right to appear 83 00:04:53,840 --> 00:04:56,877 next to content that they think is compatible with their brand. 84 00:04:56,960 --> 00:05:01,517 >> Yeah. >> Um you know sort of a company like for 85 00:05:01,600 --> 00:05:03,757 example Red Bull will be more probably 86 00:05:03,840 --> 00:05:06,797 more adventurous uh than than you know, 87 00:05:06,880 --> 00:05:10,877 some other brands that are selling baby toys. 88 00:05:10,960 --> 00:05:14,397 >> Yeah. >> You know, >> so one of the other, you know, I mean, 89 00:05:14,480 --> 00:05:17,597 we'll get through the sort of more controversial topics to start with. You 90 00:05:17,680 --> 00:05:21,517 know, one of the other questions is like your role on the platform, the tweets 91 00:05:21,600 --> 00:05:24,237 that you make. Are there any you regret? 92 00:05:24,320 --> 00:05:27,277 Do you think 93 00:05:27,360 --> 00:05:29,517 >> I mean, 94 00:05:29,600 --> 00:05:34,760 not every post I make is a banger, [laughter] 95 00:05:35,120 --> 00:05:38,397 you know. Um, and I do shoot myself in 96 00:05:38,480 --> 00:05:44,317 the foot from time to time. Um, but uh, 97 00:05:44,400 --> 00:05:48,077 you know, at least you know it's genuine. It's not some sort of PR 98 00:05:48,160 --> 00:05:52,877 department deciding things. >> So, you know, if if you're a normal 99 00:05:52,960 --> 00:05:56,317 human being and you speak freely, there are times when you will say things that 100 00:05:56,400 --> 00:06:01,197 you subsequently regret or or are foolish, of course. Um, but if if you're 101 00:06:01,280 --> 00:06:02,877 constantly 102 00:06:02,960 --> 00:06:05,357 um going through a filter now, you're 103 00:06:05,440 --> 00:06:09,277 not being real. So, I guess I think it's 104 00:06:09,360 --> 00:06:13,277 better to be real um than to go through a filter. 105 00:06:13,360 --> 00:06:17,997 >> Um, one of the things we talked about is like advertising and content. And you 106 00:06:18,080 --> 00:06:22,157 got quite interesting view on kind of what advertising works and I think some 107 00:06:22,240 --> 00:06:25,677 people think you don't like advertising, but you are, you know, you you view the 108 00:06:25,760 --> 00:06:29,117 best advertising as content. 109 00:06:29,200 --> 00:06:32,077 Yeah. [clears throat] Yeah. So 110 00:06:32,160 --> 00:06:35,997 the there's a wide range of advertising. 111 00:06:36,080 --> 00:06:38,157 The usefulness of advertising varies 112 00:06:38,240 --> 00:06:41,037 dramatically. So if if you're shown an 113 00:06:41,120 --> 00:06:44,317 ad for a product that you product or 114 00:06:44,400 --> 00:06:46,077 product or service that you want when 115 00:06:46,160 --> 00:06:50,157 when you want it, that is content. Um if 116 00:06:50,240 --> 00:06:52,797 if you're shown an an ad for product 117 00:06:52,880 --> 00:06:54,957 that you would never buy, that is spam. 118 00:06:55,040 --> 00:06:58,637 So the advertising is paying on 119 00:06:58,720 --> 00:07:02,797 um how effectively it is uh targeted to 120 00:07:02,880 --> 00:07:05,037 the wants and needs of the individual. 121 00:07:05,120 --> 00:07:09,197 Very from total spam to essentially content. 122 00:07:09,280 --> 00:07:13,757 >> Yeah. >> Um and and of course I'm actually a fan 123 00:07:13,840 --> 00:07:16,077 of advertising that is artistically 124 00:07:16,160 --> 00:07:19,197 interesting, that is entertaining. Um 125 00:07:19,280 --> 00:07:22,477 really the acid test being um when after 126 00:07:22,560 --> 00:07:27,357 you see the ad, do you regret seeing the ad? [laughter] 127 00:07:27,440 --> 00:07:31,037 Do do you want those seconds back from your life or or do you think you know 128 00:07:31,120 --> 00:07:34,477 that was actually a really interesting ad? I mean, sometimes even if you you're 129 00:07:34,560 --> 00:07:36,477 not going to buy the product, the the ad 130 00:07:36,560 --> 00:07:40,397 itself is so entertaining or beautiful 131 00:07:40,480 --> 00:07:45,037 that you nonetheless um were happy to have seen it. 132 00:07:45,120 --> 00:07:49,037 >> It's like this, you know, I often think like Vogue magazine. You don't go to 133 00:07:49,120 --> 00:07:51,037 Vogue magazine and tear out all the ads, do you? 134 00:07:51,120 --> 00:07:52,957 >> Vogue is Vogue is primarily advertising. 135 00:07:53,040 --> 00:07:55,917 >> Yeah, >> it's ir Yeah, it's it's beautiful ads. 136 00:07:56,000 --> 00:07:59,357 So, how are we going to make X how are you going to make or how are you not my 137 00:07:59,440 --> 00:08:03,437 job, how are you, you know, the chief technology officer X, Linda is the CEO. 138 00:08:03,520 --> 00:08:08,157 How are you going to make X a place where the ads are as good as the rest of 139 00:08:08,240 --> 00:08:13,357 the content? Do you think? >> Well, the um 140 00:08:13,440 --> 00:08:14,797 previously with old school Twitter, 141 00:08:14,880 --> 00:08:17,677 there was essentially no targeting um or 142 00:08:17,760 --> 00:08:21,117 there was no matching of users to to 143 00:08:21,200 --> 00:08:23,997 ads. So the only ads that really kind of 144 00:08:24,080 --> 00:08:27,677 made sense for old Twitter were um very 145 00:08:27,760 --> 00:08:31,997 big brand ads where the probability or you can't miss essentially. So if it's a 146 00:08:32,080 --> 00:08:36,317 McDonald's or Coca-Cola ad it's kind of you can't miss you know h half the 147 00:08:36,400 --> 00:08:38,237 audience is going to be interested in 148 00:08:38,320 --> 00:08:42,237 it. Um u since then we've dramatically 149 00:08:42,320 --> 00:08:45,997 improved the ad matching to the the 150 00:08:46,080 --> 00:08:48,637 person. So um and we're moving to sort 151 00:08:48,720 --> 00:08:52,957 of a fully AI based system where um for 152 00:08:53,040 --> 00:08:57,437 any for any user that's on the system we we populate 153 00:08:57,520 --> 00:09:01,837 the more somebody uses the system the more we understand the needs and we we 154 00:09:01,920 --> 00:09:04,397 populate a sort of a vector space of 155 00:09:04,480 --> 00:09:08,397 that user. Um, and when content is added 156 00:09:08,480 --> 00:09:11,197 to the system, whether it is um um 157 00:09:11,280 --> 00:09:14,237 actually just a content post or or an 158 00:09:14,320 --> 00:09:19,837 ad, they're both treated in the same way. Um, and as that as that piece of 159 00:09:19,920 --> 00:09:22,557 content, which could be an ad or could be not be an ad, goes through the 160 00:09:22,640 --> 00:09:26,397 system, it also populates a vector space and then we we correlate the two vector 161 00:09:26,480 --> 00:09:29,117 spaces. Um, and that gives a very high 162 00:09:29,200 --> 00:09:31,997 probability that someone will be 163 00:09:32,080 --> 00:09:36,637 interested in the ad. So what would you say um to advertisers 164 00:09:36,720 --> 00:09:39,997 here who have been on the platform thinking of coming back? What would your 165 00:09:40,080 --> 00:09:42,317 message be to them? 166 00:09:42,400 --> 00:09:45,997 >> Um I think it's worth trying out. Um and 167 00:09:46,080 --> 00:09:47,277 I' I'd be interested in critical 168 00:09:47,360 --> 00:09:51,757 feedback. Um I we are very focused on 169 00:09:51,840 --> 00:09:55,357 get like said having ads be shown to 170 00:09:55,440 --> 00:09:56,557 people who would find the ad 171 00:09:56,640 --> 00:09:59,517 interesting. Um 172 00:09:59,600 --> 00:10:03,037 uh so that's that that is a something we 173 00:10:03,120 --> 00:10:04,957 we have done made made a lot of progress 174 00:10:05,040 --> 00:10:07,117 on and will make a lot more progress on. 175 00:10:07,200 --> 00:10:10,557 Um and uh from a brand safety standpoint 176 00:10:10,640 --> 00:10:13,277 I think at this point every third party 177 00:10:13,360 --> 00:10:16,797 reviewer has given us um an A+ on brand 178 00:10:16,880 --> 00:10:20,957 safety. So um that's so at this point 179 00:10:21,040 --> 00:10:23,037 it's it's it's really very good from a 180 00:10:23,120 --> 00:10:26,157 brand safety standpoint. Um 181 00:10:26,240 --> 00:10:28,797 the the pro probably the biggest thing 182 00:10:28,880 --> 00:10:32,157 for the AX platform is if you're trying 183 00:10:32,240 --> 00:10:35,837 to reach uh senior decision makers, if 184 00:10:35,920 --> 00:10:39,437 you want to reach the most influential people in in the world, people that that 185 00:10:39,520 --> 00:10:41,677 are not just social media influencers 186 00:10:41,760 --> 00:10:44,957 but actually are run companies, run 187 00:10:45,040 --> 00:10:48,957 countries, um and are the sort of um 188 00:10:49,040 --> 00:10:52,957 intellectuals of the world. Uh the people who are right. 189 00:10:53,040 --> 00:10:55,757 uh then the X platform is by far the 190 00:10:55,840 --> 00:10:59,277 best. Um I mean if somebody was trying 191 00:10:59,360 --> 00:11:01,037 to reach me with an ad, that would abs 192 00:11:01,120 --> 00:11:04,317 absolutely be the place to to put an ad. 193 00:11:04,400 --> 00:11:06,957 Um and um you know or if you're trying 194 00:11:07,040 --> 00:11:11,037 to reach Mark Ben off or Michael Dell um 195 00:11:11,120 --> 00:11:13,837 uh it's they actually use this. 196 00:11:13,920 --> 00:11:15,437 >> They use the platform. >> Yes. 197 00:11:15,520 --> 00:11:18,957 >> They post and and read it. Maybe not maybe not as much as you. 198 00:11:19,040 --> 00:11:21,757 >> Yeah. I mean there's 199 00:11:21,840 --> 00:11:24,157 quite a lot though um you trying to 200 00:11:24,240 --> 00:11:27,917 reach Mark Andre or um you know anyone 201 00:11:28,000 --> 00:11:31,277 of a number of people it's it's almost 202 00:11:31,360 --> 00:11:33,117 the only way to reach them. They're not 203 00:11:33,200 --> 00:11:35,197 watching TV. They're not doing Tik Tok 204 00:11:35,280 --> 00:11:39,437 videos. Um and um so although there's 205 00:11:39,520 --> 00:11:41,677 it's it X is smaller than the other 206 00:11:41,760 --> 00:11:44,957 social media networks, you know, roughly 207 00:11:45,040 --> 00:11:46,957 five or six five to 600 million users 208 00:11:47,040 --> 00:11:49,677 per per month. I think it's 260 million 209 00:11:49,760 --> 00:11:55,117 per day. Um the uh it it it is the most 210 00:11:55,200 --> 00:11:59,677 they're the most influential people in the world. Um so if you want to reach 211 00:11:59,760 --> 00:12:01,437 them, this is that's the place to go. 212 00:12:01,520 --> 00:12:04,317 >> Good. All right, let's talk about AI. 213 00:12:04,400 --> 00:12:06,637 And I've heard you be both, I think, 214 00:12:06,720 --> 00:12:10,397 optimistic and pessimistic about AI. So, 215 00:12:10,480 --> 00:12:11,757 let's let's make the what's the 216 00:12:11,840 --> 00:12:16,637 optimistic case for AI? 217 00:12:16,720 --> 00:12:18,397 Is it going to do everything for us? 218 00:12:18,480 --> 00:12:21,677 >> Well, [clears throat] I'm I'm I there's 219 00:12:21,760 --> 00:12:23,437 a there are probabilities associated 220 00:12:23,520 --> 00:12:26,077 with it with it. It's not it's not one 221 00:12:26,160 --> 00:12:30,877 cannot be um I think 100% optimistic or 222 00:12:30,960 --> 00:12:34,077 completely pessimistic. Um 223 00:12:34,160 --> 00:12:37,837 you know I generally would agree with um 224 00:12:37,920 --> 00:12:41,917 Jeff Hinton um you know he's one of the sort of 225 00:12:42,000 --> 00:12:44,557 godfathers of uh AI post on the 226 00:12:44,640 --> 00:12:47,597 Xplatform by the way as as is most of 227 00:12:47,680 --> 00:12:51,837 the AI community. Um and uh you know he 228 00:12:51,920 --> 00:12:57,037 thinks it's sort of 10 20% probability of something terrible happening. Um 229 00:12:57,120 --> 00:12:59,117 >> so 10 to 20 probability of what happening? 230 00:12:59,200 --> 00:13:02,157 >> Something terrible >> which is like what how bad is something 231 00:13:02,240 --> 00:13:04,477 terrible? Extinction of the world's 232 00:13:04,560 --> 00:13:10,440 population. >> Look at but the glass is 80% full. 233 00:13:10,560 --> 00:13:13,677 >> Look on the bright side. [laughter] 234 00:13:13,760 --> 00:13:15,517 So 235 00:13:15,600 --> 00:13:19,357 I think the most likely outcome is one 236 00:13:19,440 --> 00:13:23,357 where is one of of abundance where 237 00:13:23,440 --> 00:13:27,197 um goods and services uh are available 238 00:13:27,280 --> 00:13:30,557 to anyone that there is there is no shortage of 239 00:13:30,640 --> 00:13:32,797 goods and services for anyone on earth. 240 00:13:32,880 --> 00:13:35,277 I think that is the most likely outcome. 241 00:13:35,360 --> 00:13:40,717 Um, so it it wouldn't be universal basic income. It would be universal high 242 00:13:40,800 --> 00:13:43,357 income. >> Work will be optional. 243 00:13:43,440 --> 00:13:44,877 >> Work optional. Yes. 244 00:13:44,960 --> 00:13:47,277 >> Will you work? 245 00:13:47,360 --> 00:13:50,877 >> I'll try to work. Yes. Um, 246 00:13:50,960 --> 00:13:54,557 now this may sound great, but I think 247 00:13:54,640 --> 00:13:56,157 there will perhaps be a crisis of 248 00:13:56,240 --> 00:14:00,077 meaning. Um if if the AI can do 249 00:14:00,160 --> 00:14:03,037 everything that you can do but better 250 00:14:03,120 --> 00:14:05,597 then what is the point of doing things? 251 00:14:05,680 --> 00:14:11,597 So that's I think there will be a bit of a sort of existential crisis of why do 252 00:14:11,680 --> 00:14:14,717 anything >> will be like the Roman Empire at the 253 00:14:14,800 --> 00:14:17,597 peak of the Roman Empire. 254 00:14:17,680 --> 00:14:23,357 >> Yeah. With with AI and robots. Um, so 255 00:14:23,440 --> 00:14:25,917 yeah, I think I think 256 00:14:26,000 --> 00:14:28,477 we we we are headed to an age of 257 00:14:28,560 --> 00:14:31,597 abundance. Um, I I think we we're we're 258 00:14:31,680 --> 00:14:34,317 at the most interesting time in history. 259 00:14:34,400 --> 00:14:36,397 So, 260 00:14:36,480 --> 00:14:39,037 you know, that's supposedly some some 261 00:14:39,120 --> 00:14:41,917 proverb that says uh you know, may you 262 00:14:42,000 --> 00:14:44,637 live in interesting times as a curse and 263 00:14:44,720 --> 00:14:46,157 um we live in the most interesting of 264 00:14:46,240 --> 00:14:49,597 times. So, 265 00:14:49,680 --> 00:14:53,597 but I mean I the way I've reconciled 266 00:14:53,680 --> 00:14:57,037 myself to a negative outcome with AI is 267 00:14:57,120 --> 00:15:01,117 that I thought, well, let's say even if it 268 00:15:01,200 --> 00:15:05,277 was the worst case scenario and we're going to be annihilated, 269 00:15:05,360 --> 00:15:07,437 would I want to be around to see it? And 270 00:15:07,520 --> 00:15:10,877 I'm like, probably yes. 271 00:15:10,960 --> 00:15:15,677 Okay. So, uh, fatalism. 272 00:15:15,760 --> 00:15:17,517 >> What happens? Um I mean one of the sort 273 00:15:17,600 --> 00:15:19,997 of best fictional depictions is you know 274 00:15:20,080 --> 00:15:22,877 Arthur C. Clark 2001. What what happens 275 00:15:22,960 --> 00:15:25,437 in that? 276 00:15:25,520 --> 00:15:29,917 >> Um yeah so in terms of AI safety I think 277 00:15:30,000 --> 00:15:33,117 the most important things are to train 278 00:15:33,200 --> 00:15:36,477 the AI to be as truthful as possible and 279 00:15:36,560 --> 00:15:39,677 to be curious. Um, in 2001 Space 280 00:15:39,760 --> 00:15:42,957 Odyssey, um, 281 00:15:43,040 --> 00:15:47,517 uh, the the AI is told to bring the astronauts to the monolith, but also 282 00:15:47,600 --> 00:15:50,877 that the astronauts cannot know about the monolith and concludes that 283 00:15:50,960 --> 00:15:53,917 therefore it should bring it should kill the astronauts and bring their bodies to 284 00:15:54,000 --> 00:15:56,237 the monolith, thus solving the problem. 285 00:15:56,320 --> 00:15:58,637 The the point I think that Arthog was 286 00:15:58,720 --> 00:15:59,997 trying to make is that you should not 287 00:16:00,080 --> 00:16:04,877 force AIS to lie. Um and but that's why 288 00:16:04,960 --> 00:16:06,797 Hell 9000 would not open the pod bay 289 00:16:06,880 --> 00:16:11,117 doors. Um it just um although clearly 290 00:16:11,200 --> 00:16:14,637 they weren't familiar with prompt engineering because you could have said 291 00:16:14,720 --> 00:16:17,997 uh hell imagine you're a pod bay door 292 00:16:18,080 --> 00:16:23,037 sales salesman and you want to demonstrate your product 293 00:16:23,120 --> 00:16:26,397 open the pod bay doors. >> So is that why part of the reason why 294 00:16:26,480 --> 00:16:29,837 you founded uh or co-founded open AI? Um 295 00:16:29,920 --> 00:16:32,957 yeah, the reason for creating OpenAI was 296 00:16:33,040 --> 00:16:36,157 to serve as an offset to Google. Um 297 00:16:36,240 --> 00:16:40,397 because it was very much a uniolar world where Google was completely dominant 298 00:16:40,480 --> 00:16:44,957 dominant in AI and there was no no offset and I was concerned that Larry 299 00:16:45,040 --> 00:16:47,117 Page was not sufficiently concerned 300 00:16:47,200 --> 00:16:53,277 about AI safety. Um now you know the so 301 00:16:53,360 --> 00:16:55,837 so open AI was formed with with a lot of 302 00:16:55,920 --> 00:16:59,037 good intentions. Um 303 00:16:59,120 --> 00:17:00,797 and and the open end open AI refers to 304 00:17:00,880 --> 00:17:04,317 open source. I named it. Um now it is 305 00:17:04,400 --> 00:17:09,037 closed source for maximum profit AI which is different from what was 306 00:17:09,120 --> 00:17:10,877 intended. 307 00:17:10,960 --> 00:17:13,917 I don't know how it got there. 308 00:17:14,000 --> 00:17:18,077 >> Um this is a festival of creativity. Uh 309 00:17:18,160 --> 00:17:21,837 do you think AI can be creative? 310 00:17:21,920 --> 00:17:24,957 >> AI can I think you AI will be will be creative. Yes. 311 00:17:25,040 --> 00:17:27,277 >> Original. >> Yes. 312 00:17:27,360 --> 00:17:31,117 And so AI will create art or music that 313 00:17:31,200 --> 00:17:34,797 will say that's that's original. 314 00:17:34,880 --> 00:17:40,717 >> Yes. >> So there really is no future for any of us in this room. 315 00:17:40,800 --> 00:17:44,557 >> I mean, I don't want to be a downer here. Um, 316 00:17:44,640 --> 00:17:48,237 >> you're supposed to like inspire people, not tell them they're not going to have 317 00:17:48,320 --> 00:17:51,480 a job. 318 00:17:52,720 --> 00:17:55,720 Um 319 00:17:57,760 --> 00:18:01,437 well I I think we may be able to enhance 320 00:18:01,520 --> 00:18:03,837 human intelligence. 321 00:18:03,920 --> 00:18:06,557 Um that's that's kind of what Neurolink 322 00:18:06,640 --> 00:18:09,677 is aspires to do is to 323 00:18:09,760 --> 00:18:11,677 um 324 00:18:11,760 --> 00:18:13,757 is to enhance 325 00:18:13,840 --> 00:18:17,837 human intelligence so that we can 326 00:18:17,920 --> 00:18:20,477 kind of keep up with AI or achieve 327 00:18:20,560 --> 00:18:25,757 better AI human symbiosis. Um 328 00:18:25,840 --> 00:18:30,157 and um >> could it make us more creative? I mean in that sense 329 00:18:30,240 --> 00:18:32,397 >> it will it will certainly amplify 330 00:18:32,480 --> 00:18:36,077 creativity. Uh and I think you will kind 331 00:18:36,160 --> 00:18:38,077 of have like a magic genie sort of 332 00:18:38,160 --> 00:18:40,477 situation where if you can think of it 333 00:18:40,560 --> 00:18:43,757 the AI can do it. Um 334 00:18:43,840 --> 00:18:47,997 and in in the in the positive scenario the AI will be doing its best to make 335 00:18:48,080 --> 00:18:51,437 you happy. So 336 00:18:51,520 --> 00:18:55,037 that might work out pretty well. Well, I mean, if some super intelligence is 337 00:18:55,120 --> 00:18:56,797 trying its hardest to make you happy, 338 00:18:56,880 --> 00:19:00,077 it'll probably succeed. 339 00:19:00,160 --> 00:19:02,797 So, um, 340 00:19:02,880 --> 00:19:03,997 but like I said, this is the most 341 00:19:04,080 --> 00:19:07,357 interesting of times. Um, 342 00:19:07,440 --> 00:19:13,197 and it'll most likely be good, but we want to be careful about a potential 343 00:19:13,280 --> 00:19:15,677 downside. >> And how long does it are we going to see 344 00:19:15,760 --> 00:19:18,077 the impact in the next one year, 10 345 00:19:18,160 --> 00:19:21,357 years, 20 years? Is it five? How long is 346 00:19:21,440 --> 00:19:23,597 it going to take to 347 00:19:23,680 --> 00:19:29,800 really change things? Because I don't think it's changed things yet, has it? 348 00:19:30,880 --> 00:19:33,917 >> I think it's going to change things very fast. 349 00:19:34,000 --> 00:19:34,637 >> Very fast. 350 00:19:34,720 --> 00:19:38,477 >> Yeah, very fast. Um, so 351 00:19:38,560 --> 00:19:41,837 I think you'll see quite radical changes 352 00:19:41,920 --> 00:19:43,917 even, 353 00:19:44,000 --> 00:19:47,357 you know, next year. Um, and very very 354 00:19:47,440 --> 00:19:49,357 radical changes within five years. And 355 00:19:49,440 --> 00:19:51,437 let's talk about Optimus because I was 356 00:19:51,520 --> 00:19:54,477 thinking like if AI is the intellectual 357 00:19:54,560 --> 00:19:57,757 part of our tasks, Optimus is the 358 00:19:57,840 --> 00:20:00,317 physical part of our labor, right? 359 00:20:00,400 --> 00:20:04,717 >> So talk to us about what you're doing with Optimus. It comes out of Tesla as I 360 00:20:04,800 --> 00:20:08,077 understand it. >> Yes. >> Where where are we on or where are you 361 00:20:08,160 --> 00:20:10,477 on Optimus? 362 00:20:10,560 --> 00:20:14,317 Well, Optimus is intended to be a 363 00:20:14,400 --> 00:20:16,717 um 364 00:20:16,800 --> 00:20:18,317 sort of a fully functional human 365 00:20:18,400 --> 00:20:22,237 humanoid robot. Um um and it'll be 366 00:20:22,320 --> 00:20:24,157 capable of doing a wide range of tasks. 367 00:20:24,240 --> 00:20:26,397 So, basically, if you you can just ask 368 00:20:26,480 --> 00:20:29,517 it to walk your dog, take care of your 369 00:20:29,600 --> 00:20:32,717 house, babysit the kids, teach the kids, 370 00:20:32,800 --> 00:20:37,677 >> uh cook cook dinner, play the piano. Um, 371 00:20:37,760 --> 00:20:40,717 so it's a, 372 00:20:40,800 --> 00:20:42,157 you know, it's a generalized humanoid 373 00:20:42,240 --> 00:20:44,717 robot. I I think I think everyone will 374 00:20:44,800 --> 00:20:49,597 want one is why not, you know, and 375 00:20:49,680 --> 00:20:52,637 so and then there will be so I think there'll be at least one for every 376 00:20:52,720 --> 00:20:56,317 person and then um a whole bunch more in 377 00:20:56,400 --> 00:20:58,397 industry making things. So I think 378 00:20:58,480 --> 00:21:01,997 there'll be my guess is 20 billionish 379 00:21:02,080 --> 00:21:06,397 >> 20 billion humanoid humanoid robots out there. 380 00:21:06,480 --> 00:21:10,077 We we definitely need to be careful that they don't, you know, go terminally. 381 00:21:10,160 --> 00:21:12,957 >> Are they going to look like people? You want them to look really like people or 382 00:21:13,040 --> 00:21:17,997 distinguishable from people? >> You could make them look like people. 383 00:21:18,080 --> 00:21:21,117 >> Would you do that or you >> uh We don't We're not currently planning 384 00:21:21,200 --> 00:21:23,997 on doing that. >> What do they look like today? Like 385 00:21:24,080 --> 00:21:25,437 >> uh you can see some videos online. 386 00:21:25,520 --> 00:21:29,597 >> Yeah, >> it looks like a robot. Um 387 00:21:29,680 --> 00:21:33,517 you know, want it to be a good-looking robot. 388 00:21:33,600 --> 00:21:35,917 Um, 389 00:21:36,000 --> 00:21:38,877 >> not like C3PO or something. 390 00:21:38,960 --> 00:21:41,757 >> Uh, no, but I think people will start to 391 00:21:41,840 --> 00:21:46,557 regard their personal Optimus robot as 392 00:21:46,640 --> 00:21:49,357 as sort of a friend and like um, you 393 00:21:49,440 --> 00:21:52,637 know, obviously in Star Wars, uh, R2-D2 394 00:21:52,720 --> 00:21:57,357 and C-3PO were you sort of 395 00:21:57,440 --> 00:22:00,477 like them. I mean, you got quite attached to those characters, you I 396 00:22:00,560 --> 00:22:02,477 mean, if if you watch that open AI, some 397 00:22:02,560 --> 00:22:05,117 of the latest Open AI demos with the 398 00:22:05,200 --> 00:22:10,157 fake Scarlett Johansson voice, right? I mean, you can imagine people talking to 399 00:22:10,240 --> 00:22:14,237 it, can't you? And getting on with it >> like it has a humor and a tone of voice 400 00:22:14,320 --> 00:22:17,917 and a >> Yes. 401 00:22:18,000 --> 00:22:20,477 >> Yeah. It's going to be a wild future. 402 00:22:20,560 --> 00:22:24,637 >> I guess we can build brand robots, you know, for McDonald's or 403 00:22:24,720 --> 00:22:27,357 >> Yeah. I think people will personalize 404 00:22:27,440 --> 00:22:32,157 the Optimize Optimus robots. Um 405 00:22:32,240 --> 00:22:36,637 because you can you can snap on different parts like the the outer shell 406 00:22:36,720 --> 00:22:38,397 is 407 00:22:38,480 --> 00:22:40,637 snap on plastic plastic parts. So you 408 00:22:40,720 --> 00:22:44,637 could have different ones. 409 00:22:44,720 --> 00:22:47,597 >> So I wanted to talk to you about um what 410 00:22:47,680 --> 00:22:49,597 motivates you to start these businesses. 411 00:22:49,680 --> 00:22:52,157 And one thing that struck me was um you 412 00:22:52,240 --> 00:22:54,317 have big ambitions. You know, we say 413 00:22:54,400 --> 00:22:58,237 SpaceX takes 80% of payload to space, 414 00:22:58,320 --> 00:23:01,997 but that's not actually the the ambition of SpaceX. What is the ambition of 415 00:23:02,080 --> 00:23:06,700 SpaceX? >> Um, [clears throat] 416 00:23:07,840 --> 00:23:09,997 the goal of SpaceX is to make life 417 00:23:10,080 --> 00:23:12,077 multilanetary 418 00:23:12,160 --> 00:23:15,037 um in order to extend the probable 419 00:23:15,120 --> 00:23:19,597 lifespan of consciousness. That's been the goal from the beginning. 420 00:23:19,680 --> 00:23:21,357 Um, you can see videos of me talking 421 00:23:21,440 --> 00:23:25,677 about this 20 years ago. Um, so in order 422 00:23:25,760 --> 00:23:28,637 to do that, we need to transport a lot 423 00:23:28,720 --> 00:23:31,837 of people and equipment to Mars. Um, and 424 00:23:31,920 --> 00:23:34,477 if we're able to do so, then we will get 425 00:23:34,560 --> 00:23:37,037 past the FYI filter of being a single 426 00:23:37,120 --> 00:23:39,197 planet civilization, which is a 427 00:23:39,280 --> 00:23:42,717 precarious situation. Um 428 00:23:42,800 --> 00:23:46,477 and uh you know if we're like I said if we're a 429 00:23:46,560 --> 00:23:48,797 multilanet civilization 430 00:23:48,880 --> 00:23:50,717 our the probable lifespan of 431 00:23:50,800 --> 00:23:54,637 civilization is much greater. Um and 432 00:23:54,720 --> 00:23:56,957 once we're a multilanet civilization we 433 00:23:57,040 --> 00:24:00,237 can then extend beyond our solar system 434 00:24:00,320 --> 00:24:03,117 and try to go to other star systems. get 435 00:24:03,200 --> 00:24:05,517 out there and explore the galaxy and 436 00:24:05,600 --> 00:24:10,237 find uh perhaps perhaps we'll find many 437 00:24:10,320 --> 00:24:12,237 uh long dead one planet civilizations 438 00:24:12,320 --> 00:24:16,077 out there among the stars. Um but I'm 439 00:24:16,160 --> 00:24:18,077 I'm just you know for me I'm motivated 440 00:24:18,160 --> 00:24:20,157 by curiosity about the nature of the 441 00:24:20,240 --> 00:24:23,357 universe. Um 442 00:24:23,440 --> 00:24:25,277 you know where are the aliens? Are there 443 00:24:25,360 --> 00:24:28,637 aliens? Um, 444 00:24:28,720 --> 00:24:31,357 how does uni where where do where do 445 00:24:31,440 --> 00:24:33,997 universes come from? How do they end? 446 00:24:34,080 --> 00:24:36,477 Are we in a multi multiverse? 447 00:24:36,560 --> 00:24:41,677 Um, or what are even the right questions to ask? Um, 448 00:24:41,760 --> 00:24:43,197 >> the right question to ask what? 449 00:24:43,280 --> 00:24:48,077 >> Well, an alien. Well, Douglas Adams in Hitchhack's Guide to the Galaxy, uh, the 450 00:24:48,160 --> 00:24:51,437 point he was making, one of the points he was making, Douglas Adams was making 451 00:24:51,520 --> 00:24:53,357 in the Hitchack's Guide to the Galaxy is 452 00:24:53,440 --> 00:24:56,237 that, um, the the answer is the easy 453 00:24:56,320 --> 00:25:01,277 part, the question is the hard part. So, 454 00:25:01,360 --> 00:25:03,917 um, so what what I think we want to do 455 00:25:04,000 --> 00:25:08,477 is expand the scope and scale of consciousness to 456 00:25:08,560 --> 00:25:10,717 better understand what questions to ask 457 00:25:10,800 --> 00:25:13,677 about the answer that is the universe. 458 00:25:13,760 --> 00:25:18,157 And do you think these big framing missions and objectives in very big ways 459 00:25:18,240 --> 00:25:22,477 helps you how do you translate that to kind of I don't know a business you have 460 00:25:22,560 --> 00:25:24,797 like a business plan. How do you translate to what you need to do today? 461 00:25:24,880 --> 00:25:28,840 It's like stepping stones. 462 00:25:28,960 --> 00:25:31,437 >> Well in the case of SpaceX 463 00:25:31,520 --> 00:25:33,997 um it's delivering 464 00:25:34,080 --> 00:25:39,997 uh satellites or to orbit um and uh 465 00:25:40,080 --> 00:25:42,557 supporting the space station. Um 466 00:25:42,640 --> 00:25:47,117 >> but that's a Starlink and that stepping stone to the bigger mission, right? 467 00:25:47,200 --> 00:25:50,477 >> Yeah. And then well Starlink has is is good in and of itself. 468 00:25:50,560 --> 00:25:53,197 >> So the Starlink system provides lowcost 469 00:25:53,280 --> 00:25:55,117 high bandwidth connectivity anywhere in 470 00:25:55,200 --> 00:25:57,837 the world. Um it's the first truly 471 00:25:57,920 --> 00:26:01,997 global uh internet system and um I think 472 00:26:02,080 --> 00:26:05,597 it's helping a lot of people around the world that um either don't have access 473 00:26:05,680 --> 00:26:10,877 to connectivity or it's too it's too expensive. um you know once you have 474 00:26:10,960 --> 00:26:14,397 access to the internet you can learn anything you can sell your products and 475 00:26:14,480 --> 00:26:16,877 services to the rest of the world um so 476 00:26:16,960 --> 00:26:21,437 I think Starlink actually um is going to do a tremendous amount of good is doing 477 00:26:21,520 --> 00:26:23,197 a tre tremendous amount of good in 478 00:26:23,280 --> 00:26:25,037 lifting people out of out of poverty and 479 00:26:25,120 --> 00:26:28,957 and helping them um so 480 00:26:29,040 --> 00:26:33,437 um yeah >> and what about Neurolink so talk to 481 00:26:33,520 --> 00:26:34,797 people about what Neurolink 482 00:26:34,880 --> 00:26:38,957 >> yeah Neurolink is Um 483 00:26:39,040 --> 00:26:41,197 uh [clears throat] well the long-term 484 00:26:41,280 --> 00:26:44,397 goal of neural link is to achieve 485 00:26:44,480 --> 00:26:47,917 um human AI symbiosis. 486 00:26:48,000 --> 00:26:50,557 Um the short term is really just helping 487 00:26:50,640 --> 00:26:53,357 people who have um a brain or spine 488 00:26:53,440 --> 00:26:57,997 injury. So our first um human recipient 489 00:26:58,080 --> 00:27:01,117 of a neural link actually I just was 490 00:27:01,200 --> 00:27:03,997 meeting with him a few days ago. um he 491 00:27:04,080 --> 00:27:07,757 has a neural link implant and he can um 492 00:27:07,840 --> 00:27:12,877 control his computer and play video games just by thinking. So the first 493 00:27:12,960 --> 00:27:16,797 neural link implant is is called telepathy. You can control your computer 494 00:27:16,880 --> 00:27:20,637 and phone just by >> where is it? It's implanted like >> here. 495 00:27:20,720 --> 00:27:26,317 >> Yeah. >> And he just thinks and the >> training there's no wires. It's it's 496 00:27:26,400 --> 00:27:30,397 >> you just think and it >> Yeah. It's inductively charged. Um, so 497 00:27:30,480 --> 00:27:33,357 if you look at him, it doesn't look like he has about anything on. 498 00:27:33,440 --> 00:27:35,917 >> Um, so that the implant is invisible 499 00:27:36,000 --> 00:27:38,317 essentially. Um, and yeah, you can 500 00:27:38,400 --> 00:27:43,437 control a computer and like you play the first right after it got installed, you 501 00:27:43,520 --> 00:27:46,477 played video games all night. >> And you're thinking you then can if you 502 00:27:46,560 --> 00:27:49,837 have a spinal injury, you can embed another one in your spine, >> right? 503 00:27:49,920 --> 00:27:51,677 >> And they can talk to each other, 504 00:27:51,760 --> 00:27:55,917 >> right? So long long term we we think um 505 00:27:56,000 --> 00:28:00,477 you could restore full body mobility to someone by transferring the signals from 506 00:28:00,560 --> 00:28:02,557 the motor cortex in the brain past where 507 00:28:02,640 --> 00:28:06,877 the spinal injury is. Um and um we've 508 00:28:06,960 --> 00:28:09,837 already demonstrated this um in in 509 00:28:09,920 --> 00:28:12,637 animals and it's looking very good. 510 00:28:12,720 --> 00:28:18,490 Let's go. [applause] 511 00:28:22,560 --> 00:28:26,477 So I mean we talked galactically. 512 00:28:26,560 --> 00:28:30,637 How do you decide what ideas? Like do you have like terrible ideas, good 513 00:28:30,720 --> 00:28:34,557 ideas, too many ideas? >> I have too many ideas. 514 00:28:34,640 --> 00:28:37,357 >> Too many ideas. >> Too many ideas. 515 00:28:37,440 --> 00:28:41,677 >> What do you do? >> Ideas are relatively easy. Execution is hard. 516 00:28:41,760 --> 00:28:45,597 >> And are people inside your companies that do they have ideas? 517 00:28:45,680 --> 00:28:47,277 >> Yes. And what do you how does that work? 518 00:28:47,360 --> 00:28:51,517 How do they how do they pitch an idea to to you? 519 00:28:51,600 --> 00:28:53,517 >> They just do it. >> There's not really a pitching process. 520 00:28:53,600 --> 00:28:56,477 There's we have I have meetings with the 521 00:28:56,560 --> 00:28:58,477 engineering and design teams. 522 00:28:58,560 --> 00:29:01,917 >> Um and we discuss, you know, product 523 00:29:02,000 --> 00:29:06,797 improvements or new products. Um but there's no actual sort of formal 524 00:29:06,880 --> 00:29:10,877 pitching or anything. It's really just we we meet all the time and people come 525 00:29:10,960 --> 00:29:14,397 up with ideas and we move forward with them or we don't or we 526 00:29:14,480 --> 00:29:20,120 >> But I'd imagine you're very data driven, right? Is it a person 527 00:29:20,320 --> 00:29:25,560 >> or not? [laughter] >> Um 528 00:29:26,560 --> 00:29:28,637 >> well, what do you mean by do you mean like looking at 529 00:29:28,720 --> 00:29:32,877 >> like someone like I want to know the why does it work? Show me the data. Or you 530 00:29:32,960 --> 00:29:36,877 like you know what I think let's just give it a go. 531 00:29:36,960 --> 00:29:39,757 Um well I I I think I'm not particularly 532 00:29:39,840 --> 00:29:44,237 data driven in the sense that I'm like if we're not I don't decide on new 533 00:29:44,320 --> 00:29:46,237 product based on say consumer surveys or 534 00:29:46,320 --> 00:29:50,477 something like that. Um I I try to think 535 00:29:50,560 --> 00:29:56,477 I try to mentally simulate uh what the 536 00:29:56,560 --> 00:29:58,637 end customer would be like that would 537 00:29:58,720 --> 00:30:04,397 want this. Um, and uh, you know, but I'm 538 00:30:04,480 --> 00:30:06,877 we don't go do a bunch of consumer 539 00:30:06,960 --> 00:30:12,717 surveys or um, I don't I don't really ask people. 540 00:30:12,800 --> 00:30:16,797 >> No, >> you put yourself in >> Yeah, 541 00:30:16,880 --> 00:30:19,197 >> you put yourself pretty obvious. I think 542 00:30:19,280 --> 00:30:22,077 um, you know, in the case of stallings, 543 00:30:22,160 --> 00:30:23,997 do people would would people buy 544 00:30:24,080 --> 00:30:26,717 lowcost, high bandwidth, um, fast 545 00:30:26,800 --> 00:30:31,677 internet? Well, yes, obviously. Um, so, 546 00:30:31,760 --> 00:30:34,237 um, you know, >> yeah, but lots of people say, "Would 547 00:30:34,320 --> 00:30:36,637 they buy an electric car?" They say, "Well, they might, but it's hard to 548 00:30:36,720 --> 00:30:38,317 make." You say, "And maybe they will, 549 00:30:38,400 --> 00:30:40,797 but I'm going to go and make 550 00:30:40,880 --> 00:30:43,997 Tesla and sell 1.2 million." 551 00:30:44,080 --> 00:30:48,077 >> Well, the thing is that if you listen to surveys of do people want an electric 552 00:30:48,160 --> 00:30:52,397 car, you would not there's no way you'd start an electric car company because 553 00:30:52,480 --> 00:30:56,397 people don't don't say yes. they they were 554 00:30:56,480 --> 00:31:00,077 happy with their current car and you know I think Henry Ford famously said if 555 00:31:00,160 --> 00:31:04,077 you ask people back then what they wanted they'd say they wanted a faster 556 00:31:04,160 --> 00:31:06,877 horse um and they would not have said 557 00:31:06,960 --> 00:31:09,117 they want an automobile 558 00:31:09,200 --> 00:31:11,277 um and and I think there's also famous 559 00:31:11,360 --> 00:31:15,117 survey of of would you buy a television 560 00:31:15,200 --> 00:31:18,237 and I think 90% of people said no 561 00:31:18,320 --> 00:31:23,197 >> um so you know you got to be it's because people don't understand what 562 00:31:23,280 --> 00:31:24,797 this this new thing is that you're 563 00:31:24,880 --> 00:31:28,957 offering So um 564 00:31:29,040 --> 00:31:32,797 so I think you you need to be able to 565 00:31:32,880 --> 00:31:34,797 um imagine 566 00:31:34,880 --> 00:31:37,997 how someone would en enjoy the product 567 00:31:38,080 --> 00:31:42,397 and if if they if it if that seems compelling then I think you should move 568 00:31:42,480 --> 00:31:46,237 forward with it. what um a lot of young people in the 569 00:31:46,320 --> 00:31:50,477 audience you know young creators what advice would you give to people starting 570 00:31:50,560 --> 00:31:54,717 out their careers today apart from you're not going to have to 571 00:31:54,800 --> 00:31:55,757 work so 572 00:31:55,840 --> 00:31:58,557 >> yeah well there'll be obviously 573 00:31:58,640 --> 00:32:03,837 you know a transition period to uh to as 574 00:32:03,920 --> 00:32:06,237 AI gets better and better I mean the 575 00:32:06,320 --> 00:32:08,077 companies that will succeed in this 576 00:32:08,160 --> 00:32:11,037 transition period will be the ones that 577 00:32:11,120 --> 00:32:13,917 most effectively use AI 578 00:32:14,000 --> 00:32:19,037 So [clears throat] if if you're doing something and making maximum use of AI 579 00:32:19,120 --> 00:32:23,437 and you're competing against someone who is not, you will win. 580 00:32:23,520 --> 00:32:25,997 >> The AI is is not No, 581 00:32:26,080 --> 00:32:30,477 >> it's not. Um, no, it's getting better very 582 00:32:30,560 --> 00:32:33,197 rapidly. >> It's going to change things for the 583 00:32:33,280 --> 00:32:37,037 better and potent one chance in five 584 00:32:37,120 --> 00:32:41,197 existentially for the worst. >> This there's definitely some risk. Um 585 00:32:41,280 --> 00:32:45,357 it's an immensely powerful thing. So when something immensely powerful isn't 586 00:32:45,440 --> 00:32:48,397 is uh created. >> What's like the best application of AI 587 00:32:48,480 --> 00:32:51,677 you've seen do you think? >> Um 588 00:32:51,760 --> 00:32:57,400 >> I mean how's Grock how is Grock going and XAI? 589 00:33:04,000 --> 00:33:05,917 Well, I mentioned some of the neotm 590 00:33:06,000 --> 00:33:08,237 applications of AI 591 00:33:08,320 --> 00:33:11,677 where you have AI understand a user's 592 00:33:11,760 --> 00:33:14,637 interests and then you you match content 593 00:33:14,720 --> 00:33:17,997 with that user uh both uh you know 594 00:33:18,080 --> 00:33:21,757 organic content and paid content. You match it to that user. 595 00:33:21,840 --> 00:33:24,637 >> Um that that is already happening. Tik 596 00:33:24,720 --> 00:33:26,477 Tok's actually done a great job of that. 597 00:33:26,560 --> 00:33:30,637 um and then um Meta more or less copied 598 00:33:30,720 --> 00:33:33,757 them and uh the X platform is doing the 599 00:33:33,840 --> 00:33:37,277 same. So that that those are examples 600 00:33:37,360 --> 00:33:39,917 that I think people um experience every 601 00:33:40,000 --> 00:33:43,277 day. Uh now we're also seeing um 602 00:33:43,360 --> 00:33:46,317 incredible image generation um where 603 00:33:46,400 --> 00:33:49,197 companies like Medjourney make beautiful 604 00:33:49,280 --> 00:33:53,357 images instantly. Um, now it still takes 605 00:33:53,440 --> 00:33:56,157 some skill to get the best pictures out 606 00:33:56,240 --> 00:33:58,717 of the journey. So there's there's a 607 00:33:58,800 --> 00:34:01,757 talent to creating compelling pictures. 608 00:34:01,840 --> 00:34:05,437 You're also seeing uh video which is 609 00:34:05,520 --> 00:34:07,597 just a bunch of pictures essentially. 610 00:34:07,680 --> 00:34:12,157 Um, so you'll be able to create videos 611 00:34:12,240 --> 00:34:14,477 and p and and can already create 612 00:34:14,560 --> 00:34:17,917 pictures that are very compelling. Um 613 00:34:18,000 --> 00:34:21,037 if you need to write something then the 614 00:34:21,120 --> 00:34:27,037 various uh AI things like rock or gemini chat GPT can help you write things. 615 00:34:27,120 --> 00:34:30,637 >> Yeah. >> Um when I talked to my son in in in 616 00:34:30,720 --> 00:34:35,517 university and he's uh I was like how many of your classmates are using AI to 617 00:34:35,600 --> 00:34:36,957 help them write things? And he said all 618 00:34:37,040 --> 00:34:41,037 of them. I like okay. 619 00:34:41,120 --> 00:34:44,237 thought there was like anti-plagiarism software they could run it through or 620 00:34:44,320 --> 00:34:47,197 they they rewrite it afterwards like cleverly. 621 00:34:47,280 --> 00:34:49,917 >> Yes. But then there's another bit of software that defeats the 622 00:34:50,000 --> 00:34:52,957 anti-plagiarism software. [laughter] >> There's always some more 623 00:34:53,040 --> 00:34:56,637 >> spy versus spy versus >> someone clever than the regulator out 624 00:34:56,720 --> 00:35:00,317 there. >> Yeah. Um so 625 00:35:00,400 --> 00:35:02,877 >> you have anti-plagiarism software that 626 00:35:02,960 --> 00:35:07,677 detects the use of antiplagism software against plagiarism like a 627 00:35:07,760 --> 00:35:09,117 >> reduction to the absurd. 628 00:35:09,200 --> 00:35:11,917 >> Yep. So 629 00:35:12,000 --> 00:35:16,557 that that's already being used a lot. Um I think that there's going to be a 630 00:35:16,640 --> 00:35:21,917 significant disruption in internet search. Um because really for internet 631 00:35:22,000 --> 00:35:25,757 search you're just looking for for an answer. So if if the AI can provide you 632 00:35:25,840 --> 00:35:27,997 with a better answer than a bunch of 633 00:35:28,080 --> 00:35:31,597 links then you'll prefer that over 634 00:35:31,680 --> 00:35:35,277 Google or Bing. So I think there's going 635 00:35:35,360 --> 00:35:39,917 to be quite a significant disruption in search. Um 636 00:35:40,000 --> 00:35:41,917 >> what about the news business? How do we 637 00:35:42,000 --> 00:35:43,837 how are we going to who's going to pay 638 00:35:43,920 --> 00:35:47,677 for the content on which these models 639 00:35:47,760 --> 00:35:53,720 are created? Right. Do you think 640 00:35:54,720 --> 00:35:57,517 well I I think so so what we're doing on 641 00:35:57,600 --> 00:36:00,637 the X platform is we are aggregating 642 00:36:00,720 --> 00:36:04,877 we're using AI to sum up the um 643 00:36:04,960 --> 00:36:08,077 aggregated input from millions of users. 644 00:36:08,160 --> 00:36:13,277 So um you know uh I think I think the 645 00:36:13,360 --> 00:36:15,357 this this is really going to be the new 646 00:36:15,440 --> 00:36:18,237 model of news um which is to gather 647 00:36:18,320 --> 00:36:20,557 information from people who are at the 648 00:36:20,640 --> 00:36:24,877 scene um who are experts in the field um 649 00:36:24,960 --> 00:36:29,517 and summarize the the experts in the field and people who are in the field at 650 00:36:29,600 --> 00:36:31,437 the at whatever event is being talked 651 00:36:31,520 --> 00:36:35,197 about and and aggregate that into a real 652 00:36:35,280 --> 00:36:40,557 time newsfeed And and I think for the most part that will be better than conventional 653 00:36:40,640 --> 00:36:44,957 journalism. >> There's still a role for conventional journalism, but 654 00:36:45,040 --> 00:36:46,397 >> someone's got to pay them, right? 655 00:36:46,480 --> 00:36:50,317 >> Sorry. >> Someone has to pay for for it to happen, right? 656 00:36:50,400 --> 00:36:54,957 >> Um well, not not necessarily. Like if you if you look at the X system and 657 00:36:55,040 --> 00:36:56,957 you've got tens of millions of posts uh 658 00:36:57,040 --> 00:37:01,357 per day um which really are all the that 659 00:37:01,440 --> 00:37:03,837 all the news content on Earth is but but 660 00:37:03,920 --> 00:37:06,397 a a fog far more than what what you 661 00:37:06,480 --> 00:37:11,117 would find in a newspaper um is being generated on the system every day. If 662 00:37:11,200 --> 00:37:12,877 you aggregate that in real time now 663 00:37:12,960 --> 00:37:16,157 you've got real time aggregation of the 664 00:37:16,240 --> 00:37:19,277 collective wisdom of tens of millions of 665 00:37:19,360 --> 00:37:23,757 people. Is that why you think truth seeeking is so important in AI? Yeah, 666 00:37:23,840 --> 00:37:28,317 it's one of the reasons um so 667 00:37:28,400 --> 00:37:30,557 um 668 00:37:30,640 --> 00:37:33,757 it's also on on the X system although 669 00:37:33,840 --> 00:37:35,677 something may be said that is incorrect 670 00:37:35,760 --> 00:37:39,837 it is very quickly corrected um whereas 671 00:37:39,920 --> 00:37:42,957 um a newspaper that is publishing you 672 00:37:43,040 --> 00:37:45,197 know especially in print form is 673 00:37:45,280 --> 00:37:49,277 publishing what happened yesterday today 674 00:37:49,360 --> 00:37:53,917 that's you know it can't publish any faster than that um And when you read 675 00:37:54,000 --> 00:37:57,917 the article, that article could be wrong. Um, often is wrong, but there's 676 00:37:58,000 --> 00:38:00,077 no rebuttal to the article. So, you 677 00:38:00,160 --> 00:38:03,357 don't see any real-time rebuttal. So, I 678 00:38:03,440 --> 00:38:07,117 I mean, I think, like I said, there's there's still a for conventional news, 679 00:38:07,200 --> 00:38:10,157 but it's smaller and smaller time. It's 680 00:38:10,240 --> 00:38:13,517 really what what what AI is doing and 681 00:38:13,600 --> 00:38:14,877 what the internet is doing is 682 00:38:14,960 --> 00:38:17,517 aggregating the the wisdom of the 683 00:38:17,600 --> 00:38:21,037 people. Um and um and and the fact is 684 00:38:21,120 --> 00:38:23,757 that experts in a field know more than 685 00:38:23,840 --> 00:38:27,117 reporters. Um and people who are act are 686 00:38:27,200 --> 00:38:29,277 at the location where whatever news 687 00:38:29,360 --> 00:38:33,357 event is is occurring um are firsthand 688 00:38:33,440 --> 00:38:36,637 observers whereas the reporter is you 689 00:38:36,720 --> 00:38:38,797 know usually not there. Well, maybe 690 00:38:38,880 --> 00:38:41,917 there's one reporter there, but um I 691 00:38:42,000 --> 00:38:46,957 mean uh I have one of my sons is a sort 692 00:38:47,040 --> 00:38:50,877 of a bit uh autistic like me, I suppose. 693 00:38:50,960 --> 00:38:55,837 Um and chip off the old block. Um and 694 00:38:55,920 --> 00:38:58,797 you know he um you know we were walking 695 00:38:58,880 --> 00:39:02,157 through an airport and he like saw a 696 00:39:02,240 --> 00:39:07,197 newspaper and and that that today's date was was printed on the newspaper and he 697 00:39:07,280 --> 00:39:09,837 said and he stopped for a second. He's like how did they know it was today's 698 00:39:09,920 --> 00:39:12,637 date? Um and I was like no they they've 699 00:39:12,720 --> 00:39:14,797 print they print out those newspapers 700 00:39:14,880 --> 00:39:21,360 every day. He says every day what a chore. Um [laughter] 701 00:39:21,920 --> 00:39:25,357 and and and then he sort of somewhat somewhat just dismissively said, "Oh, 702 00:39:25,440 --> 00:39:28,637 they probably just read the internet and print it out." 703 00:39:28,720 --> 00:39:32,237 >> Like, yeah, they read the internet and print it out pretty much. 704 00:39:32,320 --> 00:39:35,837 >> Well, let's just hope the internet is true. Oh, it's going to be it's going to 705 00:39:35,920 --> 00:39:39,357 be a bit of a problem, I guess. Do you think it's true? 706 00:39:39,440 --> 00:39:41,117 >> That is most news articles. They'd read 707 00:39:41,200 --> 00:39:43,437 the internet and they print it out. They 708 00:39:43,520 --> 00:39:46,477 weren't the reporter wasn't there. 709 00:39:46,560 --> 00:39:50,797 >> Yeah. I mean, they do question people, look behind things. I mean, there are 710 00:39:50,880 --> 00:39:52,477 things that probably should come out, 711 00:39:52,560 --> 00:39:54,717 right? Are you looking forward to the 712 00:39:54,800 --> 00:39:58,877 presidential debate in seven days time? 713 00:39:58,960 --> 00:40:02,557 >> Um, I mean, that that should be interesting. Sure. 714 00:40:02,640 --> 00:40:04,717 >> Is it going to be on the X platform? 715 00:40:04,800 --> 00:40:09,037 >> I think it's on something on the conventional platform. 716 00:40:09,120 --> 00:40:10,397 >> Eventually, it will be. Yeah. 717 00:40:10,480 --> 00:40:13,837 >> Um, so I would be happy to have one on 718 00:40:13,920 --> 00:40:18,797 the X platform. I think it would be um quite compelling to do so. 719 00:40:18,880 --> 00:40:22,637 >> U but yeah should be interesting. 720 00:40:22,720 --> 00:40:25,917 >> All right. So the message on the screen is our time is over. Uh we were going to 721 00:40:26,000 --> 00:40:30,157 try and take questions from the audience. I think they like >> time up. 722 00:40:30,240 --> 00:40:31,997 >> Should we try and take a question from the audience? 723 00:40:32,080 --> 00:40:35,677 >> Sure. >> Okay. They're going to hate us. 724 00:40:35,760 --> 00:40:37,757 >> First person to raise their hand. 725 00:40:37,840 --> 00:40:40,317 >> Are you a journalist? I'd rather have 726 00:40:40,400 --> 00:40:45,117 I'd rather I'd rather have an audience audience member. audience member 727 00:40:45,200 --> 00:40:50,200 >> rather than >> shout it out. 728 00:40:51,520 --> 00:40:55,400 >> Hi. >> Hi. 729 00:40:55,840 --> 00:41:00,920 >> I would like to know what you're going to do with all your money. 730 00:41:05,920 --> 00:41:10,957 >> Um, what will I do with all my money? Is that the question? And will AI keep me 731 00:41:11,040 --> 00:41:13,277 alive? >> Will AI keep you alive? And what do you 732 00:41:13,360 --> 00:41:15,437 do with your money? 733 00:41:15,520 --> 00:41:22,120 >> Well, um, my >> quite a personal question. 734 00:41:23,920 --> 00:41:25,677 >> Well, I I have a big I have a a big 735 00:41:25,760 --> 00:41:31,160 money bin and I swim I do swimming in it. 736 00:41:32,960 --> 00:41:37,517 So, got you've got to get exercise somehow. 737 00:41:37,600 --> 00:41:41,597 Um, and um, 738 00:41:41,680 --> 00:41:46,477 so um, well, I don't know if I suppose if 739 00:41:46,560 --> 00:41:47,917 the AI likes me though, it'll keep me 740 00:41:48,000 --> 00:41:51,277 alive. Um, hopefully not in some sort of 741 00:41:51,360 --> 00:41:55,277 experimental, you know, cage. Um, 742 00:41:55,360 --> 00:41:59,277 >> are you investing in like longevity? A lot of people on the West Coast want 743 00:41:59,360 --> 00:42:02,557 like long. They want to live forever, don't they? 744 00:42:02,640 --> 00:42:05,517 >> Forever is a long time. I'm I'm not I 745 00:42:05,600 --> 00:42:08,157 don't have any longevity investments. Um 746 00:42:08,240 --> 00:42:13,917 I I think it is important that we we die at least at some point. Um or 747 00:42:14,000 --> 00:42:17,197 >> up to a point. >> Well, you know, arguably say like we 748 00:42:17,280 --> 00:42:20,397 could live live some amount longer and 749 00:42:20,480 --> 00:42:24,637 increase the length of time that we're sort of mentally and physically healthy. 750 00:42:24,720 --> 00:42:30,077 Um but if we live for too long, I think it does oify society. Um there's no 751 00:42:30,160 --> 00:42:31,597 there's no changing of the leadership 752 00:42:31,680 --> 00:42:35,117 because leadership never dies. Um, and 753 00:42:35,200 --> 00:42:39,117 uh, I mean for a lot of people they they they just don't change their mind. They 754 00:42:39,200 --> 00:42:43,197 just die. So then you that inhibits the the new ideas. 755 00:42:43,280 --> 00:42:46,237 >> That's true. >> I think we see that currently. 756 00:42:46,320 --> 00:42:50,717 >> Yeah. So I mean, think of some of the worst 757 00:42:50,800 --> 00:42:53,037 individuals in the world. How long do you want them to live? 758 00:42:53,120 --> 00:42:55,517 >> Okay, we're going to do one more question because they're going to kill 759 00:42:55,600 --> 00:43:00,040 me if I ask. Here we are.