1 00:00:01,620 --> 00:00:06,500 But I think the development of full artificial intelligence could spell the 2 00:00:06,539 --> 00:00:08,100 end of the human race. 3 00:00:11,859 --> 00:00:16,680 Once humans develop artificial intelligence, it would take off on 4 00:00:16,699 --> 00:00:20,719 its own and redesign itself at an ever increasing rate. 5 00:00:23,100 --> 00:00:27,859 In between five and 20 years from now, there's a good chance, a 6 00:00:27,899 --> 00:00:30,839 50% chance, we'll get AI smarter than us. 7 00:00:32,680 --> 00:00:34,059 I think we're 8 00:00:34,899 --> 00:00:39,740 moving into a period when, for the first time ever, we may have 9 00:00:39,759 --> 00:00:41,339 things more intelligent than us. 10 00:00:41,340 --> 00:00:43,879 You believe they can understand? 11 00:00:43,919 --> 00:00:44,339 Yes. 12 00:00:44,360 --> 00:00:47,120 You believe they are intelligent? 13 00:00:47,279 --> 00:00:47,659 Yes. 14 00:00:47,899 --> 00:00:51,959 You believe these systems have experiences of their own 15 00:00:52,699 --> 00:00:55,419 and can make decisions based on those experiences? 16 00:00:55,860 --> 00:00:57,599 In the same sense as people do, yes. 17 00:00:57,599 --> 00:00:57,779 Are 18 00:00:58,599 --> 00:00:59,359 they conscious? 19 00:00:59,939 --> 00:01:02,639 I think they probably don't have much self-awareness at present. 20 00:01:02,999 --> 00:01:04,799 So in that sense, I don't think they're conscious. 21 00:01:04,819 --> 00:01:07,859 Will they have self-awareness, consciousness? 22 00:01:08,079 --> 00:01:08,179 Oh, yes. 23 00:01:08,179 --> 00:01:08,579 Yes? 24 00:01:08,739 --> 00:01:10,099 Oh, yes, I think they will in time. 25 00:01:10,299 --> 00:01:15,219 And so human beings will be the second most intelligent beings on the 26 00:01:15,279 --> 00:01:15,739 planet? 27 00:01:15,739 --> 00:01:16,939 Yeah. 28 00:01:17,499 --> 00:01:19,859 You believe that ChatGPT-4 29 00:01:20,639 --> 00:01:21,659 understands? 30 00:01:21,659 --> 00:01:22,319 I believe it 31 00:01:22,999 --> 00:01:24,539 definitely understands, yes. 32 00:01:24,539 --> 00:01:25,979 And in five years time? 33 00:01:25,999 --> 00:01:29,939 I think in five years time, it may well be able to reason better than us. 34 00:01:30,159 --> 00:01:32,819 And we can't afford to get it wrong with these things. 35 00:01:32,859 --> 00:01:34,759 Can't afford to get it wrong, why? 36 00:01:35,339 --> 00:01:36,759 Well, because they might take over. 37 00:01:37,759 --> 00:01:39,219 Take over from humanity? 38 00:01:39,399 --> 00:01:40,579 Yes, that's a possibility. 39 00:01:40,579 --> 00:01:41,059 Why would they want to? 40 00:01:41,059 --> 00:01:42,219 I'm not saying it will happen. 41 00:01:43,239 --> 00:01:47,339 If we could stop them ever wanting to, that would be great, but it's not clear we can 42 00:01:47,359 --> 00:01:48,619 stop them ever wanting to. 43 00:01:50,419 --> 00:01:54,039 So the question is, what's going to happen when we've created beings that are more 44 00:01:54,059 --> 00:01:54,999 intelligent than us? 45 00:01:56,099 --> 00:01:56,719 And 46 00:01:57,959 --> 00:02:01,039 we don't know what's going to happen. We've never been in that situation before. 47 00:02:01,679 --> 00:02:06,399 Anybody who says it's all going to be fine is crazy, and anybody who says they inevitably 48 00:02:06,419 --> 00:02:09,339 going to take over, they're crazy too. We really don't know. 49 00:02:09,739 --> 00:02:13,819 But because we really don't know, it will make a lot of sense to do a lot of basic 50 00:02:13,859 --> 00:02:18,719 research now on whether we can stay in control of things that we create that are 51 00:02:18,759 --> 00:02:19,879 more intelligent than us. 52 00:02:20,419 --> 00:02:23,199 What are the implications 53 00:02:24,019 --> 00:02:28,999 of these systems autonomously writing their own computer code and executing 54 00:02:29,199 --> 00:02:30,379 their own computer code? 55 00:02:30,379 --> 00:02:32,479 That's a serious worry, right? 56 00:02:33,119 --> 00:02:36,539 So one of the ways in which these systems might escape control 57 00:02:37,259 --> 00:02:40,979 is by writing their own computer code to modify themselves, 58 00:02:42,099 --> 00:02:45,079 and that's something we need to seriously worry about. 59 00:02:45,079 --> 00:02:50,019 What do you say to someone who might argue, "If the systems become malevolent, 60 00:02:50,119 --> 00:02:51,559 just turn them off?" 61 00:02:51,559 --> 00:02:54,739 They will be able to manipulate people, right? 62 00:02:55,099 --> 00:02:59,959 And these will be very good at convincing people because they'll have learned from all 63 00:02:59,979 --> 00:03:03,959 the novels that were ever written, all the books by Machiavelli, 64 00:03:04,939 --> 00:03:09,619 all the political connivances, they'll know all that stuff, they'll know how to do it. 65 00:03:13,199 --> 00:03:17,499 We thought AI would happen in decades or centuries, 66 00:03:18,339 --> 00:03:18,539 but it 67 00:03:19,419 --> 00:03:24,159 might be just in a few years. And I saw how it could go wrong because we 68 00:03:24,159 --> 00:03:26,619 didn't and we still don't have ways to make sure 69 00:03:28,239 --> 00:03:31,479 this technology eventually doesn't turn against us. 70 00:03:32,259 --> 00:03:36,979 So two months later, I'm a leading signatory of the pause 71 00:03:37,019 --> 00:03:41,619 letter where we and 30,000 other people asked the AI 72 00:03:41,679 --> 00:03:46,059 labs to wait six months before building the next version. 73 00:03:47,439 --> 00:03:49,039 As you can guess, nobody paused. 74 00:03:49,939 --> 00:03:50,339 Then, 75 00:03:51,579 --> 00:03:55,939 with the same people and the leading executives of the AI labs, 76 00:03:56,419 --> 00:04:00,279 I sign a statement, and the statement goes, 77 00:04:00,279 --> 00:04:03,659 "Mitigating the risk of extinction 78 00:04:04,299 --> 00:04:06,939 from AI should be a global priority." 79 00:04:08,739 --> 00:04:12,159 I then testify in front of the US Senate a- about those risks. 80 00:04:12,359 --> 00:04:17,299 I travel the world to talk about it. Okay, so this was a controlled experiment. 81 00:04:18,099 --> 00:04:22,499 What is it going to be in a few years when these systems are much more powerful? 82 00:04:22,719 --> 00:04:27,579 There's already studies showing that they can learn to avoid showing 83 00:04:27,619 --> 00:04:31,379 their deceptive plans in these chain of thoughts that we can monitor. 84 00:04:32,619 --> 00:04:36,999 When they'll be more powerful, they would not just copy themselves on one other computer 85 00:04:36,999 --> 00:04:38,239 and start that program. 86 00:04:38,979 --> 00:04:42,759 They would copy themselves over hundreds or thousands of computers over the internet. 87 00:04:44,279 --> 00:04:48,499 But if they really want to make sure we would never shut them down, they would have an 88 00:04:48,519 --> 00:04:51,319 incentive to get rid of us. 89 00:04:52,899 --> 00:04:53,279 So 90 00:04:54,179 --> 00:04:58,799 I know I'm asking you to make a giant leap into a future that looks so 91 00:04:58,899 --> 00:05:00,579 different from where we are now, 92 00:05:01,939 --> 00:05:04,459 but it might be just a few years away or a decade away. 93 00:05:04,899 --> 00:05:09,279 So we are on a trajectory to build machines that are smarter and smarter, 94 00:05:10,279 --> 00:05:13,939 and one day, it's very plausible that they will be smarter than us, 95 00:05:14,599 --> 00:05:14,959 and then 96 00:05:15,619 --> 00:05:17,279 they will have their own agency, 97 00:05:19,419 --> 00:05:20,419 their own goals 98 00:05:22,019 --> 00:05:25,219 which may not be aligned with ours. 99 00:05:26,659 --> 00:05:28,539 What happens to us then? 100 00:05:30,339 --> 00:05:30,939 Poof, 101 00:05:32,619 --> 00:05:33,599 we are 102 00:05:34,259 --> 00:05:36,399 blindly driving into a fog 103 00:05:37,359 --> 00:05:42,199 despite the warnings of scientists like myself that this trajectory could 104 00:05:42,259 --> 00:05:44,379 lead to loss of control. 105 00:05:45,419 --> 00:05:50,179 Beside me in the car are my children, my grandson, 106 00:05:50,459 --> 00:05:51,359 my loved ones. 107 00:05:53,379 --> 00:05:54,999 Who is beside you 108 00:05:55,639 --> 00:05:56,259 in the car? 109 00:05:58,179 --> 00:06:00,899 Who is in your care for the future? 110 00:06:00,899 --> 00:06:02,099 ... regulation. 111 00:06:02,119 --> 00:06:07,019 And- and- and just wrap it up, you know, in very clear terms, what is 112 00:06:07,039 --> 00:06:08,839 the worst case scenario that you're worried about? 113 00:06:10,410 --> 00:06:14,969 Well, the worst case scenario is, uh, human extinction, uh, 114 00:06:15,389 --> 00:06:19,910 eh, a lot of, uh, uh, the CEOs and leading AI researchers, including 115 00:06:19,929 --> 00:06:21,209 myself, Geoff Hinton, 116 00:06:21,810 --> 00:06:24,989 um, have signed a declaration saying that, 117 00:06:25,589 --> 00:06:30,009 uh, making sure we mitigate that existential risk should be a priority. 118 00:06:30,709 --> 00:06:35,370 Oh, yeah. Yes, in fact, I have reduced my expectations for when we will 119 00:06:35,449 --> 00:06:38,329 get to AGI or whatever human level. 120 00:06:39,110 --> 00:06:40,609 Uh, I used to think it would be 121 00:06:41,229 --> 00:06:44,929 5 to 20 years, and now I think it could be, you know, 2 to 10 years. 122 00:06:44,929 --> 00:06:46,549 Two years? Wow. 123 00:06:47,449 --> 00:06:47,469 Well- 124 00:06:47,469 --> 00:06:52,430 I mean, two would be the lower, look, the, the smallest but, but, uh, um, it could be 125 00:06:52,489 --> 00:06:55,690 five, it could be 10, uh, really, we should plan 126 00:06:56,350 --> 00:06:58,190 considering all these possibilities- 127 00:06:58,190 --> 00:06:58,210 Yeah. 128 00:06:58,210 --> 00:07:01,769 ... and in particular the shorter one just in case 'cause that could be catastrophic. 129 00:07:01,769 --> 00:07:06,369 And the world is ready for a potential AGI, as in human level 130 00:07:06,749 --> 00:07:07,769 AI intelligence, 131 00:07:08,949 --> 00:07:09,589 within two years? 132 00:07:09,589 --> 00:07:10,049 Not at all. 133 00:07:10,049 --> 00:07:10,709 All right. 134 00:07:10,709 --> 00:07:11,389 Not at all. 135 00:07:13,649 --> 00:07:17,209 These are his words, these are exact quotes from him, 136 00:07:17,209 --> 00:07:20,429 "Barring cataclysms, 137 00:07:21,029 --> 00:07:25,669 I consider the development of intelligent machines a near-term inevitability. 138 00:07:27,769 --> 00:07:28,589 Rather quickly, 139 00:07:29,409 --> 00:07:31,309 they would displace us from existence. 140 00:07:33,029 --> 00:07:35,029 I'm not as alarmed as many 141 00:07:36,109 --> 00:07:38,769 since I consider these future machines our progeny, 142 00:07:39,509 --> 00:07:44,269 mind children built in our image and likeness, ourselves in more potent form." 143 00:07:45,149 --> 00:07:45,969 Number three, 144 00:07:48,609 --> 00:07:51,489 this will not stop with human-level abilities. 145 00:07:51,509 --> 00:07:54,709 We will, we will move beyond human-level abilities. 146 00:07:57,609 --> 00:08:01,729 We might, uh, you might imagine we could s- we could stop but, but I think realistically 147 00:08:01,749 --> 00:08:05,869 we can't because there are many of us and point number one, there's no consensus. 148 00:08:06,349 --> 00:08:10,529 So some people will do it and it will go to the limit. 149 00:08:11,909 --> 00:08:16,849 So, and then finally over time power and resources will flow to 150 00:08:16,869 --> 00:08:20,669 the most intelligent beings, it doesn't really matter if they're human or intelligent, 151 00:08:21,169 --> 00:08:22,229 power and resources 152 00:08:22,869 --> 00:08:24,589 will flow to the most intelligent. 153 00:08:25,229 --> 00:08:25,729 So, so 154 00:08:26,429 --> 00:08:27,449 this what we just said, 155 00:08:28,649 --> 00:08:33,409 we have the transition, we have the ascent of humanity, succession to 156 00:08:33,469 --> 00:08:34,909 AI is sort of inevitable. 157 00:08:37,329 --> 00:08:39,529 So that's my second main message, 158 00:08:40,509 --> 00:08:43,609 AI, the inevitable next step in the development of universe 159 00:08:44,349 --> 00:08:48,869 and we should embrace it with courage and pride, also with humility 160 00:08:49,569 --> 00:08:51,569 (sighs) and a sense of adventure. 161 00:08:54,849 --> 00:08:59,769 And so as a result of that, I, I, and you can find it on my blog from about 162 00:08:59,789 --> 00:09:04,709 2009 I think is the first time I publicly talked about it, I predicted a 50% chance 163 00:09:04,729 --> 00:09:06,969 of, um, AGI by 2028. 164 00:09:07,669 --> 00:09:07,889 How did- 165 00:09:07,889 --> 00:09:09,609 And I still, I still believe that today. 166 00:09:09,609 --> 00:09:11,349 But, so I think that 167 00:09:12,069 --> 00:09:16,609 very, very powerful, very intelligent, um, artificial intelligence is coming. 168 00:09:17,009 --> 00:09:19,489 Um, I think theers, this is very, very likely. 169 00:09:20,049 --> 00:09:23,409 It, I don't think it's coming today, I don't think it's coming next year or the year 170 00:09:23,429 --> 00:09:26,209 after, it's probably a little bit further out than that. 171 00:09:26,209 --> 00:09:26,469 What, in 2028? You- 172 00:09:26,469 --> 00:09:27,289 And I don't think... 173 00:09:27,969 --> 00:09:32,809 2028, that's a 50% chance so, you know, if it doesn't happen in 2028 I'm not gonna be 174 00:09:32,849 --> 00:09:33,709 surprised, obviously. 175 00:09:33,709 --> 00:09:35,489 But that, but you, I think you also, 176 00:09:36,229 --> 00:09:39,969 you're not like an irredeemable optimist only, you, you see a potential- 177 00:09:40,009 --> 00:09:40,029 Yeah. 178 00:09:40,029 --> 00:09:42,149 ... for it to go very badly in a different direction. 179 00:09:42,429 --> 00:09:45,129 Talk about what that pathway could look like. 180 00:09:45,129 --> 00:09:49,789 Well, yeah. I want to explain, I think what's, I, I don't believe the people who are 181 00:09:49,869 --> 00:09:53,769 sure that it's gonna go very well and I don't believe the people who are sure that it's 182 00:09:53,789 --> 00:09:58,189 gonna go very, very badly because what we're talking about is an incredibly 183 00:09:58,289 --> 00:10:02,969 profound transition. It's like the arrival of human intelligence in the 184 00:10:02,989 --> 00:10:07,269 world. This is another intelligence arriving in the world and so it is an 185 00:10:07,269 --> 00:10:12,029 incredibly deep transition and we do not fully understand all the 186 00:10:12,029 --> 00:10:13,929 implications and consequences of this. 187 00:10:14,929 --> 00:10:19,229 But there is, you know, many things we don't understand about the future so I have to 188 00:10:19,269 --> 00:10:24,209 accept that there is a possibility that things m- may go badly because 189 00:10:24,329 --> 00:10:28,309 I don't know what's gonna happen. I, I can't know that about the future and such a big 190 00:10:28,369 --> 00:10:28,689 change 191 00:10:29,349 --> 00:10:34,329 and even if the probability of something going bad is quite small, we 192 00:10:34,389 --> 00:10:36,449 should take this extremely seriously. 193 00:10:36,469 --> 00:10:36,849 But then- 194 00:10:36,969 --> 00:10:41,909 Yeah. If I had a magic wand and I could slow things down, I would, I would use 195 00:10:41,909 --> 00:10:46,449 that magic wand but I don't. There are dozens of company- well, there's probably 10 196 00:10:46,489 --> 00:10:50,489 companies in the world now that can develop the most cutting edge, uh, models, 197 00:10:51,109 --> 00:10:55,369 uh, including, I think, some national, um, intelligence services who have secret 198 00:10:55,409 --> 00:10:55,969 projects. 199 00:10:56,989 --> 00:11:01,909 How, how do you stop this process? I don't see any way, realistic plan that 200 00:11:01,949 --> 00:11:03,529 I've heard of, of stopping this process. 201 00:11:03,529 --> 00:11:06,969 And so you have i- i- it's interesting 'cause you have these, uh, blog posts that you 202 00:11:06,989 --> 00:11:08,809 wrote when you started DeepMind- 203 00:11:09,009 --> 00:11:09,109 Yeah. 204 00:11:09,109 --> 00:11:11,769 ... um, you know, back in 2008 where you talk about, 205 00:11:12,409 --> 00:11:15,849 um, th- the motivation was, uh, to accelerate safety. 206 00:11:15,909 --> 00:11:16,329 Mm-hmm. 207 00:11:16,369 --> 00:11:20,569 O- on net what do you think the impact of DeepMind has been on safety versus 208 00:11:20,589 --> 00:11:21,449 capabilities? 209 00:11:21,469 --> 00:11:22,929 Ooh. 210 00:11:24,209 --> 00:11:24,909 Interesting. 211 00:11:26,409 --> 00:11:27,689 (sighs) 212 00:11:30,049 --> 00:11:32,349 I don't know. It's hard to d- hard to judge actually. 213 00:11:32,889 --> 00:11:36,509 Reasoning process and I think we're now at that first part, I think we can start training 214 00:11:36,549 --> 00:11:40,649 models now with the scale of the data is beyond what a human can experience in a 215 00:11:40,649 --> 00:11:43,029 lifetime. So I think this is the first unlocking step 216 00:11:44,169 --> 00:11:48,709 and so, yeah, I think there's a 50% chance that I'm saying by 217 00:11:48,869 --> 00:11:52,889 2028. Now it, it's just a 50% chance. I mean, I'm, I'm sure what's gonna happen is we're 218 00:11:52,889 --> 00:11:56,369 gonna get to, you know, 2029 and someone's gonna say, "Oh Shane, you were wrong." It's 219 00:11:56,409 --> 00:11:58,149 like, "Come on, it's a 50% chance." 220 00:11:58,789 --> 00:11:59,069 (laughs) So- 221 00:12:05,849 --> 00:12:10,129 One of my motivations in creating OpenAI 222 00:12:10,489 --> 00:12:12,889 was, in addition to developing this technology- 223 00:12:13,725 --> 00:12:18,586 ... was also to address the questions that are posed by AGI, the difficult questions, 224 00:12:18,586 --> 00:12:20,385 the concerns that we raised. 225 00:12:21,305 --> 00:12:25,845 In addition to working with governments and helping them understand what is coming and 226 00:12:25,865 --> 00:12:30,746 prepare for it, we also doing a lot of research on addressing the technological 227 00:12:30,845 --> 00:12:34,525 side of things so that the AI will never want to go rogue. 228 00:12:35,265 --> 00:12:37,526 And this is something which I'm working on as well. 229 00:12:38,685 --> 00:12:42,606 But still, if you, if you extrapolate this kind of progress, we'll be in a very, very 230 00:12:42,705 --> 00:12:43,585 different place 231 00:12:44,425 --> 00:12:46,125 in five years, let alone 10 years. 232 00:12:47,505 --> 00:12:49,285 It doesn't seem implausible. 233 00:12:49,905 --> 00:12:53,425 It doesn't seem at all implausible that people have computers, 234 00:12:54,626 --> 00:12:57,965 data centers that are much smarter than people. 235 00:12:58,205 --> 00:13:02,505 And by smarter, I don't mean just have more memory or have more knowledge, but I also ha- 236 00:13:02,565 --> 00:13:05,385 mean have deeper insight into 237 00:13:06,185 --> 00:13:09,706 the same subjects that we people are studying and looking into. 238 00:13:10,045 --> 00:13:13,666 It means learn even faster than people. 239 00:13:14,705 --> 00:13:16,405 Like, what could such AIs do? 240 00:13:17,166 --> 00:13:22,145 I don't know. Certainly, if such an AI were the basis of some artificial life, it would 241 00:13:22,185 --> 00:13:22,485 be... 242 00:13:23,145 --> 00:13:27,245 Well, how do you even think about it? If you have some very powerful data center that's 243 00:13:27,245 --> 00:13:28,125 also alive 244 00:13:28,805 --> 00:13:30,946 in a sense, that's what we are talking about. 245 00:13:31,426 --> 00:13:35,026 And when I imagine this world, I w- I, my, my reaction is, "Gosh, this is very 246 00:13:35,045 --> 00:13:37,565 unpredictable what's gonna happen." Very unpredictable. 247 00:13:38,105 --> 00:13:41,266 But the bare minimum, there is a bare minimum which we can articulate, 248 00:13:42,965 --> 00:13:43,325 that 249 00:13:44,805 --> 00:13:45,405 if such 250 00:13:46,265 --> 00:13:50,485 super... If, if such very, very intelligent, super intelligent data centers are built, 251 00:13:50,485 --> 00:13:51,245 being built at all, 252 00:13:52,065 --> 00:13:56,846 we want those data centers to hold warm and positive feelings towards 253 00:13:56,985 --> 00:13:58,686 people, towards humanity. 254 00:13:59,725 --> 00:14:01,265 Because those... This is going to be 255 00:14:02,265 --> 00:14:03,185 non-human life, 256 00:14:04,105 --> 00:14:07,185 in a sense. Potentially, it could be, potentially be that. 257 00:14:08,105 --> 00:14:11,285 And so I would want that any instance of such 258 00:14:13,105 --> 00:14:15,266 super intelligence to have warm feelings towards humanity. 259 00:14:15,545 --> 00:14:17,025 And so this is what we are doing with 260 00:14:17,785 --> 00:14:22,066 the Superalignment Project. Yes, the, the, the friendship piece, I think is optional, but 261 00:14:22,185 --> 00:14:24,325 I do think that we want a very pro-social AI. 262 00:14:25,265 --> 00:14:27,265 I think it's... I think it's possible. 263 00:14:27,265 --> 00:14:29,005 I don't think it's guaranteed, but I think it's possible. 264 00:14:29,005 --> 00:14:29,706 If 265 00:14:36,025 --> 00:14:41,585 you 266 00:14:41,625 --> 00:14:46,345 extrapolate the curves that we've had so far, right? If, if you say, "Well, I don't know. 267 00:14:46,345 --> 00:14:50,805 We're starting to get to, like, PhD level, and, and last year we were at undergraduate 268 00:14:50,845 --> 00:14:54,865 level, and the year before, we were at, like, the level of a high school student." Again, 269 00:14:54,886 --> 00:14:57,966 you can, you can quibble with at what tasks and for what. 270 00:14:58,345 --> 00:15:00,706 We're still missing modalities, but those are being added. 271 00:15:00,706 --> 00:15:03,826 Like, computer use was added, like, image generation has been added. 272 00:15:03,985 --> 00:15:08,085 If you just kind of, like, eyeball the rate at which these capabilities are increasing, 273 00:15:08,865 --> 00:15:12,825 it does make you think that we'll get there by 2026 or 2027. 274 00:15:13,045 --> 00:15:16,105 I think there's still worlds where it doesn't happen in, in a 100 years. 275 00:15:16,505 --> 00:15:19,705 Those wor- the number of those worlds is rapidly decreasing. 276 00:15:19,885 --> 00:15:24,645 We are rapidly running out of truly convincing blockers, truly compelling 277 00:15:24,645 --> 00:15:26,245 reasons why this will not happen in... 278 00:15:27,405 --> 00:15:31,945 Yes. I mean, I think this is... Yeah. This is, this is one of the core things that, you 279 00:15:31,965 --> 00:15:36,425 know, whether it's the, the bio risks from models that, you know, I've kind of stated in 280 00:15:36,425 --> 00:15:41,245 testimony, you know, are likely to come in two to three years, or the risks 281 00:15:41,285 --> 00:15:45,165 from truly autonomous models, which I think are more than that, but might not be a whole, 282 00:15:45,205 --> 00:15:46,885 whole lot more than that. 283 00:15:47,185 --> 00:15:47,385 Yeah. 284 00:15:47,885 --> 00:15:52,825 I would, again, emphasize the testing and auditing regime for all the risks ranging 285 00:15:52,845 --> 00:15:57,785 from, you know, those, those we f- we face today, like misinformation came up, to 286 00:15:57,785 --> 00:16:02,245 the biological risks that I'm worried about in two or three years, to the, uh, you know, 287 00:16:02,305 --> 00:16:06,225 risks of autonomous replication that are some unspecified period after. 288 00:16:06,985 --> 00:16:10,425 Um, but, but I often put it in a simple way, which is, one, the systems are getting much 289 00:16:10,465 --> 00:16:11,065 more powerful. 290 00:16:11,685 --> 00:16:12,185 Two, 291 00:16:12,885 --> 00:16:17,545 we... There, there's obviously not much of a barrier to getting the systems to 292 00:16:17,625 --> 00:16:19,405 act autonomously in the world, right? 293 00:16:19,405 --> 00:16:24,105 People have taken GPT-4, for instance, and turned it into AutoGPT. 294 00:16:24,105 --> 00:16:28,985 There was even a WormGPT, which is, you know, s- supposed to act as a computer worm. 295 00:16:29,305 --> 00:16:33,525 So, you know, powerful smart systems that can take action. 296 00:16:33,545 --> 00:16:34,685 So, you know, kind of 297 00:16:35,565 --> 00:16:38,125 very long leash of su- of human supervision. 298 00:16:38,485 --> 00:16:41,785 And because of the way they're trained, they're not easy to control. 299 00:16:41,785 --> 00:16:46,285 I mean, we all saw Bing in Sydney. So you put those three things together, and, you know, 300 00:16:46,285 --> 00:16:47,265 there's at least, 301 00:16:47,985 --> 00:16:52,405 uh, you know, I think some chance that as the systems get more and more powerful, 302 00:16:53,265 --> 00:16:57,905 you know, they're, they're gonna do things that we don't want them to do, and it may be, 303 00:16:57,965 --> 00:17:01,745 it may be difficult to fully control them, to fully rein them humans. 304 00:17:01,945 --> 00:17:05,045 And there's a further thing of, like, can it get past humans? 305 00:17:05,305 --> 00:17:10,225 Can it in- can it outperform the sum total of human, say, you know, 306 00:17:10,305 --> 00:17:13,645 scientific or engineering output? That's, that's like a... 307 00:17:13,645 --> 00:17:15,285 You know, that's, that's another point. 308 00:17:15,585 --> 00:17:20,145 That point could be, you know, could be like a year away because the model gets... 309 00:17:20,185 --> 00:17:23,645 Is better at making itself smarter and smarter, or it could be many years away. 310 00:17:23,905 --> 00:17:27,225 And then there's this further point of like, okay, you know, could the model, like, you 311 00:17:27,245 --> 00:17:31,065 know, like, explore the universe and send out a bunch of, like, you know, von Neumann 312 00:17:31,105 --> 00:17:34,665 probes, and, you know, build Dyson spheres around the sun, and, you know, 313 00:17:35,605 --> 00:17:40,565 calculate the meaning of life is 42 or whatever. Um, you know, that's, that's like a... 314 00:17:40,565 --> 00:17:44,265 That's, that's like a further point that also raises questions about, you know, what's 315 00:17:44,625 --> 00:17:48,425 practical in an engineering sense in, in all of these kind of weird, weird things. 316 00:17:48,425 --> 00:17:50,445 So that's, that's like another further point. 317 00:17:50,785 --> 00:17:54,725 It's possible all of these points are pretty compressed together because there's like a 318 00:17:54,745 --> 00:17:57,285 feedback loop, but it's possible they're very far away from... 319 00:17:57,285 --> 00:18:02,225 You know, I think I've, I think I've often said that, you know, my, my chance that 320 00:18:02,325 --> 00:18:07,293 something goes...... you know, really quite catastrophically wrong on the scale 321 00:18:07,333 --> 00:18:09,053 of, of, you know, human civilization. 322 00:18:09,053 --> 00:18:13,493 You know, it might be somewhere between 10 and 25% when you put 323 00:18:13,514 --> 00:18:17,474 together the risk of something going wrong with the model itself 324 00:18:18,073 --> 00:18:22,893 with, you know, something going wrong with human, you know, people or organizations or 325 00:18:22,974 --> 00:18:27,793 nation states misusing the model or, or it kind of inducing conflict 326 00:18:27,853 --> 00:18:32,453 among them or, or just some way in which kind of society can't, can't handle it at the 327 00:18:32,493 --> 00:18:36,353 end of it. And if you were to pause for an indefinite length of time, then you raise 328 00:18:36,373 --> 00:18:38,793 these questions like, "Well, h- how do you really get everyone to stop? 329 00:18:38,793 --> 00:18:40,154 There's an international system here. 330 00:18:40,353 --> 00:18:43,774 There's dictators who wanna use this stuff to take over the world." I mean, you know, 331 00:18:43,793 --> 00:18:47,693 people use that as an excuse, but it's also true. Uh, quite, quite what to do. 332 00:18:47,734 --> 00:18:52,433 I think our concern was that with the rate at which the technology was progressing, 333 00:18:52,853 --> 00:18:57,373 a kind of big, loud public release might accelerate things so 334 00:18:57,493 --> 00:19:01,793 fast that the ecosystem might not, might not, um, know how to handle it. 335 00:19:01,793 --> 00:19:06,493 And, you know, I didn't want our kind of first act on the public stage, you know, after 336 00:19:06,613 --> 00:19:07,793 we'd said e- you know, after we 337 00:19:08,553 --> 00:19:13,313 put so much effort into bein', bein', bein', uh, responsible to, to accelerate 338 00:19:13,373 --> 00:19:15,493 things, to accelerate things so greatly. 339 00:19:15,973 --> 00:19:19,353 Um, I, you know, certainly as, you know, as soon, as soon as the other models were out 340 00:19:19,353 --> 00:19:22,693 and kind of the gun had been fired, then we started putting these things out. 341 00:19:22,693 --> 00:19:26,213 We're like, "Okay. All right. You know, n- now, now there's definitely a market in this. 342 00:19:26,213 --> 00:19:30,153 People, people, people, people know about it, and so, you know, we should, we should, we 343 00:19:30,213 --> 00:19:33,893 should, we should, we should get out ahead." And, you know, indeed, we've, we've managed 344 00:19:33,993 --> 00:19:37,793 to, you know, put ourselves among, you know, among the top two or three players in this 345 00:19:37,813 --> 00:19:42,013 space. With... and, you know, it, it, y- y- it, it really created an environment where 346 00:19:42,013 --> 00:19:44,353 things were, you know, racing forward very quickly. 347 00:19:44,353 --> 00:19:47,413 And look, I, I love technology as much as the next person. 348 00:19:47,413 --> 00:19:51,253 There was something like, you know, super exciting about the whole, you know, "Mak- make 349 00:19:51,313 --> 00:19:53,413 them dance. Oh, we're responding with something." You know? 350 00:19:53,413 --> 00:19:57,153 I mean, I can, I can get just as excited about this as everyone, but given the rate at 351 00:19:57,153 --> 00:20:01,253 which the technology is progressing, um, you know, there was a worrying aspect about this 352 00:20:01,273 --> 00:20:05,593 as well. Um, and so i- in this case, I'm at least, at least on balance glad that, you 353 00:20:05,613 --> 00:20:07,693 know, we, we weren't the ones who fired that starting gun. 354 00:20:07,733 --> 00:20:08,313 Yeah, got it. 355 00:20:08,313 --> 00:20:08,613 ... 356 00:20:11,473 --> 00:20:12,873 about this or we're about that. 357 00:20:13,073 --> 00:20:18,013 And what was the thought at the time of, "Hey, we should go do something 358 00:20:18,173 --> 00:20:21,593 on our own"? W- was that y- your thought? Was it a group of people? 359 00:20:21,753 --> 00:20:24,093 Yeah, it was, it was definitely the thoughts of a group of people. 360 00:20:24,153 --> 00:20:29,113 So there were seven co-founders who left, um, and then I, I think in total we got 361 00:20:29,173 --> 00:20:33,753 four- 14 or 15 people from OpenAI, which was about 10% of the, the size of the 362 00:20:33,793 --> 00:20:37,913 organization at the time. Um, it w- you know, it was, it's funny to look back on those 363 00:20:37,953 --> 00:20:42,773 days because, you know, in those days, we were the language model part of OpenAI. 364 00:20:42,833 --> 00:20:47,673 Like, we, w- you know, we along with a couple people who stayed, uh, you know, were, 365 00:20:47,693 --> 00:20:52,013 were those who had developed and scaled the, the language models. 366 00:20:52,013 --> 00:20:54,933 There were many different other things going on at OpenAI, right? 367 00:20:54,933 --> 00:20:59,013 There was, you know, there was a robotics project, a theorem proving project, projects to 368 00:20:59,033 --> 00:21:01,533 play video games. Some of those still exist, 369 00:21:02,193 --> 00:21:06,613 um, but, uh, you know, we felt that we were this kind of coherent group of people. 370 00:21:06,913 --> 00:21:08,753 We had this view about, 371 00:21:09,373 --> 00:21:13,213 you know, language models and scaling, which to be fair I think the organization 372 00:21:13,293 --> 00:21:18,193 supported, but then we also had this view about, you know, we, we need to, we need 373 00:21:18,193 --> 00:21:23,033 to make these models safe in a certain way, and, you know, we need to do them 374 00:21:23,033 --> 00:21:27,953 within an organization where we can really believe that these principles are incorporated 375 00:21:28,013 --> 00:21:28,593 top to bottom. 376 00:21:28,593 --> 00:21:33,393 What's, what's something foundational, uh, I guess not to use a cute 377 00:21:33,433 --> 00:21:33,673 term- 378 00:21:33,673 --> 00:21:33,713 (laughs) 379 00:21:33,713 --> 00:21:38,433 ... but something foundational that you believe at Anthropic that you would draw in 380 00:21:38,453 --> 00:21:39,753 distinction to 381 00:21:40,393 --> 00:21:41,533 others in this space? 382 00:21:41,533 --> 00:21:44,633 Yeah. I would, I would say a couple things. 383 00:21:44,833 --> 00:21:49,773 Um, so, you know, one is just this idea that we should be building in 384 00:21:50,273 --> 00:21:52,253 safety from the beginning with models. 385 00:21:52,253 --> 00:21:57,233 And, and in particular, safety is itself a task, like, you know, is 386 00:21:57,253 --> 00:22:00,013 this thing, the model's doing right or wrong? We're, we're, we're... 387 00:22:00,013 --> 00:22:03,913 You know, right or wrong could be something as prosaic as, is the model telling you how 388 00:22:03,913 --> 00:22:08,873 to hotwire a car or as, you know, scary and sophisticated as, is the model gonna 389 00:22:08,893 --> 00:22:13,013 help me build, y- build a bio weapon or is it gonna, you know, take over the world and 390 00:22:13,053 --> 00:22:16,293 make swarms of nanobots or, you know, whatever futuristic thing? 391 00:22:16,713 --> 00:22:20,933 Figuring out whether the model's gonna do that and the behavior of the model is itself an 392 00:22:20,973 --> 00:22:23,373 intellectual task of the kind that models do. 393 00:22:23,813 --> 00:22:24,253 And so 394 00:22:25,293 --> 00:22:29,493 the problem and the solution to the problem are mixed together in this way. 395 00:22:32,573 --> 00:22:35,053 You know, it's quite a dangerous, 396 00:22:35,953 --> 00:22:37,173 quite a dangerous technology, 397 00:22:37,953 --> 00:22:39,153 um, and I, I, 398 00:22:39,853 --> 00:22:40,953 I fear I may have 399 00:22:41,673 --> 00:22:43,653 done some things to accelerate it, which is... 400 00:22:44,833 --> 00:22:45,293 I don't know. 401 00:22:46,833 --> 00:22:47,113 So 402 00:22:48,053 --> 00:22:51,633 I mean, I guess, y- you know, part of what I've been fighting and maybe what has slowed 403 00:22:51,633 --> 00:22:53,953 me down somewhat is that I'm a, I- I'm a little... 404 00:22:53,973 --> 00:22:55,893 I don't wanna, I don't wanna make Terminator real. 405 00:22:55,973 --> 00:22:56,213 (laughs) 406 00:22:56,213 --> 00:22:56,773 (laughs) 407 00:22:56,773 --> 00:22:57,933 You know? So 408 00:22:58,713 --> 00:23:00,393 I- I've been sort of, I guess, 409 00:23:01,113 --> 00:23:06,053 at least until recent years, dragging my feet on, on AI and ro- and 410 00:23:06,093 --> 00:23:07,113 humanoid robotics. 411 00:23:08,213 --> 00:23:11,573 And then I sort of come to the realiz- realization it's, it's happening whether I do it 412 00:23:11,593 --> 00:23:12,353 or not, so 413 00:23:13,393 --> 00:23:17,813 you got really two choices. Particip- y- you can either be a spectator or a participant. 414 00:23:18,673 --> 00:23:21,953 And so, like, well, I guess I'd rather be a, a participant than a spectator. 415 00:23:23,053 --> 00:23:27,293 Um, so now it's, you know, pedal to the metal on humanoid robots and, um, 416 00:23:28,793 --> 00:23:29,993 digital superintelligence. 417 00:23:32,973 --> 00:23:35,793 Like, I think, I think we're quite close to digital superintelligence. 418 00:23:37,013 --> 00:23:38,053 I think it may happen 419 00:23:39,013 --> 00:23:42,053 this year, and if it doesn't happen this year, next year for sure. 420 00:23:42,693 --> 00:23:44,633 A digital superintelligence defined as 421 00:23:45,533 --> 00:23:50,353 smarter than any human at anything. Um, I, I guess I sort of agree with Geoff 422 00:23:50,353 --> 00:23:53,393 Hinton that maybe it's a 10 to 20% chance of annihilation. 423 00:23:53,893 --> 00:23:57,733 Uh, but look on the bright side, that's 80 to 90% probability of a great outcome. 424 00:23:57,733 --> 00:24:01,213 The other conversation we've had, and 425 00:24:01,913 --> 00:24:06,733 you came out in the same side as Geoffrey Hinton in this, was 426 00:24:07,353 --> 00:24:12,101 80% of probability it's gonna be awesome.20% probability we're 427 00:24:12,101 --> 00:24:12,642 screwed. 428 00:24:12,881 --> 00:24:12,901 Yes. 429 00:24:12,901 --> 00:24:14,621 Are you still- are you still on those odds? 430 00:24:14,621 --> 00:24:19,341 Yeah. Yeah, I mean, like I say, it's- it's- it's- 431 00:24:19,521 --> 00:24:23,581 it's most likely gonna be great, and there's- there's some chance, which could be 10 to 432 00:24:23,621 --> 00:24:24,642 20% that, 433 00:24:25,261 --> 00:24:27,441 um, it goes bad. Um, 434 00:24:28,521 --> 00:24:30,101 the chance is not zero that it goes bad, 435 00:24:31,341 --> 00:24:34,181 but overall, at one point, one could say that the cup is 80% full, 436 00:24:34,962 --> 00:24:37,441 uh, is one positive way to look at it. Maybe 90%. 437 00:24:39,401 --> 00:24:44,301 That's not even a question for me, whether we're gonna go take a swing at building 438 00:24:44,321 --> 00:24:47,181 the next thing. I'm just incapable of not doing that. 439 00:24:47,461 --> 00:24:50,722 There's a- I would guess that, like, sometime in the next 12 to 440 00:24:51,261 --> 00:24:56,141 we'll reach the point where, like, most of the code that's going towards 441 00:24:56,141 --> 00:25:00,281 these efforts is written by AI. And I don't- as- as this goes on. 442 00:25:00,341 --> 00:25:03,201 This isn't like that one to two year thing of what happens when you have a- 443 00:25:03,941 --> 00:25:05,822 a- like a super powerful software engineer. 444 00:25:06,222 --> 00:25:06,461 But 445 00:25:07,241 --> 00:25:09,681 I think over time, you know, if- if you... 446 00:25:09,681 --> 00:25:13,201 Like, everyone is gonna have these superhuman tools, um, to be able to create a ton of 447 00:25:13,201 --> 00:25:15,441 different stuff. I think you're gonna get this incredible diversity. 448 00:25:15,901 --> 00:25:17,701 Um, part of it is gonna be solving, 449 00:25:18,641 --> 00:25:23,541 like, things that we hold up as, like, these, like, hard problems, like 450 00:25:23,801 --> 00:25:26,061 solving diseases or, like, 451 00:25:27,161 --> 00:25:32,081 solving different things around science or, um, or just, like, different technology that 452 00:25:32,101 --> 00:25:33,581 makes our lives better. But 453 00:25:34,441 --> 00:25:36,961 I would guess that a lot of it is going to end up being 454 00:25:37,821 --> 00:25:41,681 kind of cultural and social pursuit and enter- Basically, okay, there's security holes in 455 00:25:41,701 --> 00:25:46,241 so many different things, and if you could travel back in time a year or two 456 00:25:46,281 --> 00:25:47,521 years, right? It's like, 457 00:25:48,241 --> 00:25:51,361 that's not AI. It's like you just... Let's say you just have like one year or two years 458 00:25:51,421 --> 00:25:55,381 more knowledge of the security holes, you could pretty much hack into, like, any system, 459 00:25:55,541 --> 00:25:59,261 right? So like, you know, I stay up at night more worrying, well, what if, like, 460 00:25:59,981 --> 00:26:04,341 some actor that... Whatever. It's like from wherever you sit, there's gonna be some actor 461 00:26:04,381 --> 00:26:07,761 who you don't trust. If they're the ones who have, like, the super strong AI, whether 462 00:26:07,781 --> 00:26:12,321 it's some, like, other government that we- that does sort of like an opponent of- of- of 463 00:26:12,341 --> 00:26:16,601 our country or some company that you don't trust or whate- whatever it is, um, 464 00:26:17,721 --> 00:26:17,961 like, 465 00:26:18,721 --> 00:26:20,361 I think that that's potentially 466 00:26:21,341 --> 00:26:26,181 a much bigger risk. As in- I- I do think that there- there is some 467 00:26:26,241 --> 00:26:27,961 set of people who are afraid of like, 468 00:26:28,781 --> 00:26:32,421 you know, it really just kind of spins and goes from being like somewhat intelligent to 469 00:26:32,441 --> 00:26:35,341 extremely intelligent overnight. And I just think that there's all these physical 470 00:26:35,361 --> 00:26:38,301 constraints that make that, so that that's unlikely to happen. 471 00:26:38,801 --> 00:26:42,061 Um, I- I- I just don't, I don't really see that- that playing out. 472 00:26:42,641 --> 00:26:46,201 I know there- there's all this science fiction about, okay, you create intelligence and 473 00:26:46,621 --> 00:26:51,421 now it like starts taking on all these human-like behaviors and- and things like that. 474 00:26:51,421 --> 00:26:55,261 But I actually think that the current incarnation of all this stuff at least kind of 475 00:26:55,281 --> 00:26:58,981 feels like it's going in a direction where intelligence can be pretty separated from 476 00:26:59,041 --> 00:27:03,201 consciousness and agency and things like that, that, um, 477 00:27:03,921 --> 00:27:06,721 I think just makes it a super valuable tool, if I had to guess. 478 00:27:07,101 --> 00:27:11,941 It's- it's really interesting to me that you basically agree with the premise that there 479 00:27:11,961 --> 00:27:15,461 will be an intelligence explosion and something like super intelligence on either end. 480 00:27:15,721 --> 00:27:17,861 But then if that's the case, tell me if I'm misunderstanding you. 481 00:27:17,901 --> 00:27:21,281 If that's the case, why even bother with personal assistants and whatever? 482 00:27:21,281 --> 00:27:24,501 Why not just get- get to super human intelligence first and then deal with everything 483 00:27:24,501 --> 00:27:24,541 else later? 484 00:27:24,541 --> 00:27:27,401 Well, I think that that's just one aspect of the flywheel. 485 00:27:28,021 --> 00:27:28,381 Right? So 486 00:27:29,281 --> 00:27:34,121 part of what- what I generally disagree with on the fast takeoff thing is it takes time 487 00:27:34,161 --> 00:27:36,041 to build out physical infrastructure. 488 00:27:36,041 --> 00:27:36,221 Mm-hmm. 489 00:27:36,381 --> 00:27:41,221 Right? So if you want to build like a gigawatt cluster of compute, 490 00:27:42,021 --> 00:27:44,081 that just is gonna take some time. 491 00:27:47,021 --> 00:27:49,441 (music) 492 00:27:51,281 --> 00:27:55,561 Stargate, put that name down in your books, because I think you're going to hear a lot 493 00:27:55,621 --> 00:27:59,881 about it in the future. A new American company that will invest $500 494 00:27:59,901 --> 00:28:03,021 billion at least in AI infrastructure. 495 00:28:03,681 --> 00:28:05,781 The data centers are actually under construction. 496 00:28:05,901 --> 00:28:08,141 The first of them are under construction in Texas. 497 00:28:08,141 --> 00:28:11,041 The Abilene location, which is our first location. 498 00:28:11,801 --> 00:28:15,561 Abilene is a town on the western central plains of Texas. 499 00:28:15,921 --> 00:28:18,881 We kind of scratch and fight for everything good that comes our way. 500 00:28:18,881 --> 00:28:23,181 I had people reaching out and they said, "When is the President going to come see you?" 501 00:28:23,481 --> 00:28:26,841 And I said, "Well, he hasn't texted me yet, so..." (laughs) 502 00:28:26,841 --> 00:28:29,181 I'm thrilled we get to do this in the United States of America. 503 00:28:29,461 --> 00:28:32,081 I think this will be the most important project of this era. 504 00:28:32,081 --> 00:28:36,941 Last month I came to, um, celebrate your winning 505 00:28:37,321 --> 00:28:41,461 and promised that, uh, we will invest $100 billion. 506 00:28:41,881 --> 00:28:46,241 And you told me, "Oh, Masa, go for 200." (laughs) 507 00:28:47,581 --> 00:28:49,241 Now I came back with 500. 508 00:28:49,301 --> 00:28:53,521 Explain the math to us. How does it all add up to needing $500 billion? 509 00:28:53,521 --> 00:28:57,721 Well, that covers the capacity we think we need for the next few years, given our growth 510 00:28:57,741 --> 00:29:01,621 projections. The- the interesting question is, if we had- if we knew how to get a 511 00:29:01,621 --> 00:29:03,141 trillion dollars right now, which we don't, 512 00:29:03,781 --> 00:29:06,961 would we be able to deploy that profitably in the next few years? 513 00:29:08,481 --> 00:29:12,281 And I'm not sure about that, but I feel confident we can, like, make 500 billion of value 514 00:29:12,301 --> 00:29:12,601 back. 515 00:29:13,081 --> 00:29:17,521 I was in the real estate business. These buildings, these are big, beautiful buildings, 516 00:29:17,521 --> 00:29:18,941 they're going to employ a lot of people. 517 00:29:18,941 --> 00:29:23,141 How do you personally grapple with the pace of it all? 518 00:29:23,261 --> 00:29:25,641 It does feel like it's going very fast. 519 00:29:25,861 --> 00:29:29,881 Certainly think if I could like transport myself back three years ago, it would seem like 520 00:29:30,161 --> 00:29:31,581 unimaginable progress. 521 00:29:31,581 --> 00:29:36,381 The money piece, I think so long as AI is delivering 522 00:29:36,501 --> 00:29:37,921 value to society, 523 00:29:38,541 --> 00:29:42,961 the money will be there to build the infrastructure. I have no concerns around that. 524 00:29:43,441 --> 00:29:46,801 Um, even if it's not all 100% accounted for today, 525 00:29:47,401 --> 00:29:51,261 that money is there, and there's a lot of money that wants to invest in this category. 526 00:29:51,941 --> 00:29:55,941 What we're undergoing right now is sort of the largest infrastructure build in human 527 00:29:55,961 --> 00:30:00,401 history, right? This is like where we built the interstate high- highway system. 528 00:30:00,861 --> 00:30:05,721 You know, that enabled massive interstate commerce and transportation and catalyzed this 529 00:30:05,761 --> 00:30:07,441 massive boom in the US economy. 530 00:30:08,481 --> 00:30:12,861 What's happening in AI infrastructure is we're building the superhighway system that's 531 00:30:12,861 --> 00:30:16,741 going to enable modern intelligence systems, that's going to create all sorts of new job 532 00:30:16,821 --> 00:30:18,401 categories that don't even exist today. 533 00:30:18,461 --> 00:30:18,661 Mm-hmm. 534 00:30:18,661 --> 00:30:21,281 It's going to create new ways of doing business, it's going to create new ways of 535 00:30:21,301 --> 00:30:24,541 invention, um, and it's going to really drive humanity forward. 536 00:30:24,541 --> 00:30:28,781 I don't think the world has really had the humanoid robots moment yet. 537 00:30:29,161 --> 00:30:31,341 Like the first day, you're like walking down the street 538 00:30:32,061 --> 00:30:35,241 and there's like seven robots that walk past you doing things or whatever, it's going to 539 00:30:35,361 --> 00:30:36,361 feel very sci-fi. 540 00:30:37,201 --> 00:30:40,421 And I don't think that's very far away from like a visceral, like, oh man, this is going 541 00:30:40,421 --> 00:30:41,961 to do a lot of things that people used to do. 542 00:30:41,981 --> 00:30:46,461 So- so yeah, it's coming, and we- we have always tried to just be super honest about what 543 00:30:46,481 --> 00:30:50,141 we think the impact may be, realizing that we'll be wrong on a lot of details. 544 00:30:50,141 --> 00:30:53,261 It doesn't exist. 545 00:30:53,261 --> 00:30:57,461 How confident are you that AI's promise will outweigh the perils? 546 00:30:57,461 --> 00:31:01,501 I'm no- I'm not confident, but the... 547 00:31:02,081 --> 00:31:04,521 You know, that's- that's what we need to get oriented around. 548 00:31:04,561 --> 00:31:08,421 That's why I was surprised AI wasn't really an issue in the last election. 549 00:31:08,501 --> 00:31:10,341 It's definitely got to be shaped. 550 00:31:10,341 --> 00:31:13,201 Hearing you say you're not confident is kind of scary. 551 00:31:13,201 --> 00:31:14,221 People should 552 00:31:15,021 --> 00:31:19,681 think through, how is AI going to reshape, uh, our lives? 553 00:31:19,761 --> 00:31:24,621 It is- it is a profound, uh, change agent. 554 00:31:24,641 --> 00:31:28,741 And, you know, I don't think it should just be the people who know the technology, 555 00:31:28,761 --> 00:31:32,201 because there's a lot of choices about how it gets used. It will...