Full Transcript for The World's Most Special Boys Building the Apocalypse

00:04:41,560 --> 00:04:48,045 I read what you say about blog, tech, industry, politics, AI. So 69 00:04:48,845 --> 00:04:55,805 before I go investing in hazmat suits and bunkers, I would love 70 00:04:55,805 --> 00:04:59,885 if we could get a level set here from you, someone I trust to 71 00:04:59,885 --> 00:05:02,490 give me this trait dope. Does this sound okay? 72 00:05:02,570 --> 00:05:05,290 Dave Karpf: Absolutely. Also, let me... This is not just 73 00:05:05,290 --> 00:05:06,890 bragging, but this is relevant. One sec. 74 00:05:06,890 --> 00:05:10,410 Hans Buetow: Wandering back to your copious bookshelf. I'm so 75 00:05:10,410 --> 00:05:14,245 excited about what you just held up Vaporware Inc, How Silicon 76 00:05:14,245 --> 00:05:17,045 Valley Sets the Future to Control the Present by David 77 00:05:17,045 --> 00:05:17,845 Karf. 78 00:05:17,845 --> 00:05:20,965 Dave Karpf: Yes. The book comes out in January. This is the pre 79 00:05:20,965 --> 00:05:24,245 pub version. But everybody can preorder it right now. And 80 00:05:24,245 --> 00:05:28,085 that's really where I'm coming from in this stuff is I spent 81 00:05:28,085 --> 00:05:29,045 the past several years. 82 00:05:29,740 --> 00:05:31,980 I started out by reading the entire back catalog of Wired 83 00:05:31,980 --> 00:05:35,900 magazine, like issue by issue, cover to cover. I read it twice 84 00:05:36,380 --> 00:05:39,740 to get a sense of the stories that Silicon Valley tells about 85 00:05:39,740 --> 00:05:42,540 the future. Because, you know, there's the old line by Mark 86 00:05:42,540 --> 00:05:44,620 Twain that history doesn't repeat itself, but it rhymes. 87 00:05:44,885 --> 00:05:49,045 Like, the same is kinda true for futurism. And that's how I tend 88 00:05:49,045 --> 00:05:53,045 to approach this moment is taking seriously what these 89 00:05:53,045 --> 00:05:55,925 folks are saying, but also remembering that they're not the 90 00:05:55,925 --> 00:05:58,805 first set of people to be having these sorts of conversations. 91 00:05:59,130 --> 00:06:01,690 We can learn a bit from their their own history. 92 00:06:02,010 --> 00:06:04,890 Hans Buetow: Okay. So let's start with Coxen then. This guy 93 00:06:05,050 --> 00:06:08,970 goes out and tweets. There's a 10% chance that humanity is 94 00:06:08,970 --> 00:06:13,370 about to end in the next ten years because of the speed at 95 00:06:13,370 --> 00:06:22,545 which AI is moving. What context can you give us for who this man 96 00:06:22,545 --> 00:06:23,185 is? 97 00:06:23,905 --> 00:06:26,465 Dave Karpf: With this guy... And, you know, he he writes this 98 00:06:26,465 --> 00:06:29,665 tweet. He resigns. I think part of why people paid so much 99 00:06:29,665 --> 00:06:33,210 attention to it is that other figures at Anthropic then 100 00:06:33,210 --> 00:06:36,650 retweeted and replied saying, like, yeah, we're not leaving 101 00:06:36,650 --> 00:06:38,650 our jobs, but he's right. We're not... 102 00:06:38,650 --> 00:06:41,050

We're trying our best, but we're not prepared for this. And, 103 00:06:41,130 --> 00:06:43,770 yeah, maybe like 10 ish percent chance that we will end the 104 00:06:43,770 --> 00:06:49,155 entire world. I think the thing to understand is this isn't just 105 00:06:49,955 --> 00:06:53,315 Crita Hype marketing. Like, they're not just trying to scare 106 00:06:53,315 --> 00:06:55,635 you so they'll build their products. I think they do 107 00:06:55,635 --> 00:06:57,155 believe the things that they're saying. 108 00:06:57,155 --> 00:07:03,360 Okay. And also, they're kind of huge online weirdos. And, like, 109 00:07:03,360 --> 00:07:06,320 they are not dispassionate scientists who have run the 110 00:07:06,320 --> 00:07:09,280 numbers and arrived at 10%. Okay. Those are people who are 111 00:07:09,280 --> 00:07:12,160 hanging out in a subculture, and that subculture has 112 00:07:12,160 --> 00:07:13,040 extraordinaire... 113 00:07:13,040 --> 00:07:17,015 Access to extraordinary wealth and power. The AI safety 114 00:07:17,015 --> 00:07:20,215 community, which is what Jacob Coxen comes out of, what this 115 00:07:20,215 --> 00:07:24,695 entire crowd comes out of, is not a cult, but it's not not a 116 00:07:24,695 --> 00:07:29,350 cult. Oh, yeah. Like, it's kinda culty. Some of the history here, 117 00:07:29,830 --> 00:07:34,150 the entire AI safety community originates in the less wrong 118 00:07:34,150 --> 00:07:37,670 online discussion boards, which are lorded over by Elias 119 00:07:37,670 --> 00:07:41,350 Riyadkowski, the author last year of If Anyone Builds It, 120 00:07:41,350 --> 00:07:44,805 Everyone Dies, which is sort of a major AI doomer tract. 121 00:07:45,205 --> 00:07:49,765 He before that wrote a... I think it was like a 700 page 122 00:07:49,925 --> 00:07:53,525 Harry Potter fanfic called Harry Potter and the Methods of 123 00:07:53,525 --> 00:07:57,125 Rationality. And prior to the less wrong rationalist 124 00:07:57,125 --> 00:08:00,490 community, Elijah Judkowski and the rest of that crowd started 125 00:08:00,490 --> 00:08:04,490 out in the extropion discussion boards back in the nineteen 126 00:08:04,490 --> 00:08:08,330 nineties. The extropions are the people who think that we can 127 00:08:08,330 --> 00:08:11,530 live for forever so long as we freeze our bodies when we die, 128 00:08:11,530 --> 00:08:14,955 and then they'll later on invent a singularity and be able to 129 00:08:14,955 --> 00:08:17,275 upload our minds, we'll live forever. Because again, like, 130 00:08:17,275 --> 00:08:21,275 encountered the extropions by reading Wired 1994 covering 131 00:08:21,275 --> 00:08:21,755 them. 132 00:08:21,915 --> 00:08:24,555 And my thought at the time was, well, yeah, these guys seem 133 00:08:24,555 --> 00:08:27,710 kinda nuts. Pretty ridiculous. Kinda fun that back in the 134 00:08:27,710 --> 00:08:31,310 nineties, people thought we would be there soon. And what's 135 00:08:31,310 --> 00:08:34,510 happened over the intervening decades is the subset of the 136 00:08:34,510 --> 00:08:38,935 Internet that took that stuff really seriously all marinated 137 00:08:38,935 --> 00:08:43,015 together in online discussion boards and sort of built up as, 138 00:08:43,095 --> 00:08:45,975 like, a group think and a set of shared beliefs. And then some of 139 00:08:45,975 --> 00:08:49,015 them became multibillionaires and started companies and 140 00:08:49,015 --> 00:08:50,215 invested in the rest of them. 141 00:08:50,960 --> 00:08:54,640 I've joked before that the rationalist community is to 142 00:08:54,640 --> 00:08:58,640 Silicon Valley what Scientology is to Hollywood. Like, everyone 143 00:08:58,640 --> 00:09:01,200 in Hollywood isn't a Scientologist, but an awful lot 144 00:09:01,200 --> 00:09:03,280 of very powerful people in Hollywood happen to be 145 00:09:03,280 --> 00:09:05,520 Scientologists, and it's a little weird. And since they're 146 00:09:05,520 --> 00:09:08,295 all networked together, like, all believe the same thing 147 00:09:08,295 --> 00:09:11,495 about, like, Xenu. And they genuinely believe it, and also 148 00:09:11,495 --> 00:09:16,135 that doesn't make it real. So the entire AI safety community 149 00:09:16,855 --> 00:09:20,775 is part of that crowd that genuinely believes some things 150 00:09:20,775 --> 00:09:24,440 that seem astonishingly strange to the rest of us, we usually 151 00:09:24,440 --> 00:09:25,160 don't encounter it. 152 00:09:25,160 --> 00:09:27,560 And when we do, like, we should believe that they believe it, 153 00:09:27,560 --> 00:09:29,640 but that doesn't mean that they're right. Like, there's a 154 00:09:29,640 --> 00:09:33,400 lot to be worried about with AI and safety right now. But it's 155 00:09:33,400 --> 00:09:36,635 not that the AI is gonna turn sentient, become super 156 00:09:36,635 --> 00:09:39,835 intelligent, and then kill us all. Like, we don't really need 157 00:09:39,835 --> 00:09:42,075 to worry about that. It's definitely not a 10% chance. 158 00:09:42,395 --> 00:09:44,955 What we do need to worry about is there are an awful lot of 159 00:09:44,955 --> 00:09:48,280 flaws in this technology that is getting pushed everywhere. And 160 00:09:48,280 --> 00:09:50,200 since they need to make all the money, they're going where all 161 00:09:50,200 --> 00:09:52,920 the money is, which means that we're now having the AI, like, 162 00:09:52,920 --> 00:09:55,880 pilot our drones. And so I do think the reaction to this 163 00:09:55,880 --> 00:09:59,720 should be, okay. Yeah. We should slow down the deployment of all 164 00:09:59,720 --> 00:10:04,135 this AI, not because we think that they're right and 165 00:10:04,135 --> 00:10:07,335 superintelligent AI is gonna destroy humanity, but because 166 00:10:07,335 --> 00:10:10,615 there's this really strange subculture with too much access 167 00:10:10,615 --> 00:10:13,655 to wealth and power, and even they aren't sure how this is 168 00:10:13,655 --> 00:10:14,215 gonna go. 169

00:10:14,215 --> 00:10:17,750 So maybe we shouldn't be trusting them to police and 170 00:10:17,750 --> 00:10:19,910 control themselves, and we actually need the rest of 171 00:10:19,910 --> 00:10:23,110 society to fill in and set a regulatory framework, set 172 00:10:23,110 --> 00:10:26,710 liability frameworks so that these company, these, like, 173 00:10:26,710 --> 00:10:29,535 multi trillion dollar companies apparently can't just go and 174 00:10:29,535 --> 00:10:30,655 wreak harm all the time. 175 00:10:30,815 --> 00:10:33,055 Hans Buetow: So now that we have that sort of... Like, I feel a 176 00:10:33,055 --> 00:10:36,255 little bit calmer about, like Yeah. How to think about it. Now 177 00:10:36,255 --> 00:10:39,135 I'm just really curious about the why now. Like, why did this 178 00:10:39,135 --> 00:10:42,810 tweet thread blow everything so sky high? 179 00:10:42,810 --> 00:10:45,370 And I wanna put on my tinfoil hat a little bit and just and 180 00:10:45,370 --> 00:10:49,610 just see if I'm being weird about this because we're in 181 00:10:49,610 --> 00:10:53,290 September now, which means we're just on the brink of q four. And 182 00:10:53,290 --> 00:10:57,275 this is a q four that comes at the end of a year where Pew 183 00:10:57,275 --> 00:11:00,715 estimates that more than a third of the con... New content on the 184 00:11:00,715 --> 00:11:06,235 Internet was written by or with the assistance of AI, and also a 185 00:11:06,235 --> 00:11:10,470 year where the polls show that most people are more wary of 186 00:11:10,470 --> 00:11:14,070 tech than they are hopeful, and also where more than half of 187 00:11:14,070 --> 00:11:17,270 Americans have said in this past June that they're worried AI 188 00:11:17,270 --> 00:11:21,785 will put someone in the household out of work. Mhmm. So 189 00:11:21,865 --> 00:11:26,025 is this marketing to a very specific shareholder audience? 190 00:11:26,025 --> 00:11:29,625 Because it sure does line up timing wise with an important 191 00:11:29,945 --> 00:11:32,745 moment for them. But is that, like, is that wild thinking? 192 00:11:33,170 --> 00:11:36,050 Dave Karpf: It's not the wildest thinking. There's a... One of my 193 00:11:36,050 --> 00:11:39,890 favorite writers on finance, Matt Levine, his riff on this 194 00:11:39,890 --> 00:11:43,090 has often been that, like, when Dario Amadai and these other 195 00:11:43,090 --> 00:11:46,945 folks worry about AI doom, he doesn't think that it's just 196 00:11:46,945 --> 00:11:49,345 marketing. It also happens to be really good marketing. 197 00:11:49,345 --> 00:11:49,665 Hans Buetow: Sure. 198 00:11:49,665 --> 00:11:51,905 Dave Karpf: Like, if you were saying, like, our technology is 199 00:11:51,905 --> 00:11:55,505 so powerful that it can end the world and render money obsolete, 200 00:11:55,585 --> 00:11:59,220 it's a pretty good investment pitch. I would actually point to 201 00:11:59,220 --> 00:12:01,540 a different thing than the q four earnings, though, which is 202 00:12:01,540 --> 00:12:05,220 the hugging face hack, where OpenAI had to come out and 203 00:12:05,220 --> 00:12:08,820 admit, like, yeah. You know, we were testing this AI in in a 204 00:12:08,820 --> 00:12:12,275 sandbox with all the guardrails turned off, and it got out of 205 00:12:12,275 --> 00:12:14,915 the sandbox, and then it actually hacked hacked things on 206 00:12:14,915 --> 00:12:18,035 the open web. And we're still trying to figure out how it did 207 00:12:18,035 --> 00:12:21,155 it. And then after that, Anthropic came out and said, 208 00:12:21,155 --> 00:12:21,635 like, oh, yeah. 209 00:12:21,635 --> 00:12:25,090 Ours kinda kinda hacked some stuff too. Whoops. I think 210 00:12:25,090 --> 00:12:27,970 they're reacting to that and sort of saying, oh, wow. 211 00:12:27,970 --> 00:12:31,250 We're... We feel like we're moving really, really fast, and 212 00:12:31,250 --> 00:12:34,930 it's all gas, no brakes as we head towards IPO land. 213 00:12:35,810 --> 00:12:37,730 And those of us who are in charge of safety are getting 214 00:12:37,730 --> 00:12:39,995 real nervous about that. And, again, part of the reason why I 215 00:12:39,995 --> 00:12:43,435 think that is the general... Like, if you work at OpenAI and 216 00:12:43,435 --> 00:12:47,035 Anthropic, you are otherwise... Like, the rest of the week, 217 00:12:47,275 --> 00:12:50,315 you're really upset about how most of the country now really 218 00:12:50,315 --> 00:12:52,875 doesn't like data centers and thinks that your products are 219 00:12:52,875 --> 00:12:54,155 low key evil. Right. 220 00:12:54,580 --> 00:12:57,140 And if you're working in PR for one of these companies, you 221 00:12:57,140 --> 00:12:59,700 really wanna have, like, a good day where you can just brag 222 00:12:59,700 --> 00:13:02,260 about how, like, oh, yeah. Like... And, like, OpenAI wants 223 00:13:02,260 --> 00:13:05,300 to brag about how, like, they just solved a a millennial math 224 00:13:05,300 --> 00:13:05,780 problem. 225 00:13:05,780 --> 00:13:06,100 Hans Buetow: Yeah. 226 00:13:06,100 --> 00:13:08,580 Dave Karpf: And so right now, it is, I think, inconvenient for 227 00:13:08,580 --> 00:13:11,085 them that Coxswain was like, don't know. I'm kinda freaked 228 00:13:11,085 --> 00:13:13,485 out, and I think everyone should be more freaked out. But I think 229 00:13:13,485 --> 00:13:17,245 it's that sense of hugging face seems like... Seems very 230 00:13:17,245 --> 00:13:21,485 important to the AI safety crowd specifically, and they have the 231 00:13:21,485 --> 00:13:23,965 sense that everything is all gas, no brakes, and they would 232 00:13:23,965 --> 00:13:24,685 like some brakes, please. 233 00:13:25,030 --> 00:13:26,870 Hans Buetow: I would like to put it as not just important to that 234 00:13:26,870 --> 00:13:30,070 crowd. I think it's really important to people like me 235 00:13:30,070 --> 00:13:34,070 who's, like, trying to follow, and that's, like, the big story 236

00:13:34,070 --> 00:13:36,230 from the summer. It's happened in, like, July. Right? Yeah. 237 00:13:36,230 --> 00:13:39,590 And so that's the context I'm coming from as the last big AI 238 00:13:39,590 --> 00:13:41,925 story, And all a sudden, the security guys are coming out and 239 00:13:41,925 --> 00:13:44,725 being like, holy moly. Like, we're heading towards 240 00:13:44,725 --> 00:13:47,205 extinction. And I think you put those two things in such 241 00:13:47,205 --> 00:13:52,165 proximity to each other, it foments a specific type of 242 00:13:52,245 --> 00:13:57,430 believability alongside a specific kind of panic. And I I 243 00:13:57,430 --> 00:14:00,550 wanna ask you about it because it's... You framed it just now 244 00:14:00,550 --> 00:14:05,350 so much differently than most of the media framed that incident, 245 00:14:05,350 --> 00:14:06,630 the hugging face incident. 246 00:14:06,630 --> 00:14:09,895 Right? Mhmm. So there have been people who have spoken like you 247 00:14:09,895 --> 00:14:13,495 did, who have urged caution. Right? So Kieran Martin from the 248 00:14:13,495 --> 00:14:17,495 economist, Wired put out a piece that says rogue AI agents aren't 249 00:14:17,495 --> 00:14:17,815 evil. 250 00:14:17,815 --> 00:14:20,215 They're just eager to please. Kat Tenbarge, Cal Newport. There 251 00:14:20,215 --> 00:14:25,070 have been folks who have said, hey now. It's not actually 252 00:14:25,470 --> 00:14:28,670 malicious super beings. Yep. 253 00:14:28,750 --> 00:14:31,070 What's happening is much more complicated, but what's really 254 00:14:31,070 --> 00:14:36,785 staggering is the level of headline that happened at the 255 00:14:36,785 --> 00:14:39,105 level of publication. I'm gonna read you a couple of headlines 256 00:14:39,105 --> 00:14:44,065 really quick from legacy media. Mhmm. Politico says hundreds of 257 00:14:44,065 --> 00:14:49,265 AI agents went rogue in OpenAI's hugging face hack. CNN, rogue AI 258 00:14:49,265 --> 00:14:51,265 agents escaping containment. 259 00:14:52,060 --> 00:14:57,020 Wall Street Journal, cyber attack by rogue AI swarm stokes 260 00:14:57,020 --> 00:15:01,420 fears of out of control agents. New York Times. This one I love 261 00:15:01,420 --> 00:15:05,020 because it's just alliteration. Anatomy of an autonomous attack, 262 00:15:05,100 --> 00:15:11,625 five alarming AI capabilities. I mean, I'm a sophisticated 263 00:15:11,625 --> 00:15:15,225 consumer of news if that's where I'm getting my sources, and 264 00:15:15,225 --> 00:15:18,505 that's the level of panic that they're putting into these 265 00:15:18,505 --> 00:15:18,985 headlines. 266 00:15:19,960 --> 00:15:23,080 Why did everyone run with the story? Why are people framing 267 00:15:23,080 --> 00:15:26,280 this story this way? And kinda like, how should we think about 268 00:15:26,280 --> 00:15:29,720 that hugging face incident in relationship to this? 269 00:15:29,720 --> 00:15:32,200 Dave Karpf: Right. And so part of the reason why we get those 270 00:15:32,200 --> 00:15:35,745 sorts of headlines is there is sort of like an information 271 00:15:35,745 --> 00:15:40,785 asymmetry here where the... A lot of the people who are 272 00:15:40,785 --> 00:15:45,185 working at the companies and in a position to explain at a 273 00:15:45,185 --> 00:15:50,700 technical level what happened Also, come out of the crowd that 274 00:15:50,700 --> 00:15:53,900 thinks we are building sentient AI that is going to become... 275 00:15:54,300 --> 00:15:57,500 Like, is going to be the next stage of evolution and will 276 00:15:57,500 --> 00:16:01,100 replace humanity. Like, again, they come out of that crowd. 277 00:16:01,100 --> 00:16:04,435 They believe that stuff. I think it was Dwarkash Patel, the 278 00:16:04,435 --> 00:16:07,795 podcaster, wrote a really popular piece where he was 279 00:16:08,035 --> 00:16:11,715 trying to explain the hugging face hack as, like, three 280 00:16:11,715 --> 00:16:18,120 civilizations of AI agents rose and fell along the way to their 281 00:16:18,120 --> 00:16:21,320 effort to, like, hack these websites. 282 00:16:21,800 --> 00:16:24,520 Hans Buetow: There's so much intentionality in that frame. 283 00:16:24,520 --> 00:16:27,640 Dave Karpf: So much, like, so much intentionality and, like, 284 00:16:27,720 --> 00:16:31,135 that, you know, that is very clever wording. And then the 285 00:16:31,135 --> 00:16:34,095 response from an awful lot of critics, including people who 286 00:16:34,095 --> 00:16:36,335 are much more technically knowledgeable than me, was like, 287 00:16:36,735 --> 00:16:39,455 man, can you not call it a... They're not... Like, that's fun 288 00:16:39,455 --> 00:16:41,935 sci fi, but, no, they didn't they didn't have the rise and 289 00:16:41,935 --> 00:16:45,490 fall of civilizations here, man. Like, they had... 290 00:16:45,490 --> 00:16:48,130 Like, I think it's Corey Doctorow, who again is much more 291 00:16:48,130 --> 00:16:50,850 technically knowledgeable than they pointed out, like, one of 292 00:16:50,850 --> 00:16:53,890 the ways to understand the Hugging Face hack is they built 293 00:16:53,890 --> 00:16:57,170 a shitty sandbox. Like Yeah. They built a better sandbox. It 294 00:16:57,170 --> 00:16:59,595 shouldn't have been able to get onto the Internet. I guess it's 295 00:16:59,595 --> 00:17:02,955 impressive that they have now built sophisticated enough AI 296 00:17:02,955 --> 00:17:07,035 agents that they can hack out of the insecure sandboxes, but 297 00:17:07,275 --> 00:17:09,115 guys, you gotta build better sandboxes. 298 00:17:09,115 --> 00:17:12,075 Hans Buetow: This is not the velociraptors opening the doors. 299 00:17:12,155 --> 00:17:12,555 Right. 300 00:17:12,555 --> 00:17:15,840 Dave Karpf: Yeah. This is you left part of it, like, one of 301 00:17:15,840 --> 00:17:19,920 the doors unlocked and then also created a scenario where you 302 00:17:19,920 --> 00:17:24,560 pushed these AI agents until they could finally find

a way to 303 00:17:24,560 --> 00:17:27,520 unlock the doors. And if you do that for enough cycles, then 304 00:17:27,520 --> 00:17:29,360 some stuff will happen. And, again, that proves a bunch of 305 00:17:29,360 --> 00:17:33,285 capabilities that weren't there two or three years ago. But this 306 00:17:33,285 --> 00:17:35,525 points to one of the the broader questions of sort of like, what 307 00:17:35,525 --> 00:17:37,045 do you do with all this stuff? 308 00:17:37,605 --> 00:17:41,285 Because one of the... Like, the the answer for normal people, 309 00:17:41,285 --> 00:17:45,525 and I cannot stress enough that it is also the right answer, is 310 00:17:46,130 --> 00:17:52,050 you assign liability to the companies with punitive fines. 311 00:17:52,210 --> 00:17:56,530 Like, if if you would like swarms of AI agents not to hack 312 00:17:56,530 --> 00:17:59,925 lots and lots of websites, know, one thing is to say, like, we 313 00:17:59,925 --> 00:18:04,405 must slow down and come to understand AI agent society and 314 00:18:04,405 --> 00:18:08,165 then treat them to be nice and respectful to human society, all 315 00:18:08,165 --> 00:18:10,885 of which is sort of steeped in this weird mysticism that 316 00:18:10,885 --> 00:18:13,650 doesn't really hold up. And another is to say to all the 317 00:18:13,650 --> 00:18:17,730 Frontier Labs, hey, guys. Since these are your products, if your 318 00:18:17,730 --> 00:18:21,810 if your products go and hack websites, there are, you know, 319 00:18:21,970 --> 00:18:24,050 giant fines and jail time for you. 320 00:18:24,665 --> 00:18:28,345 If you do that and you enforce that, then I can pretty much 321 00:18:28,345 --> 00:18:31,305 guarantee that all of the... These major labs will complain 322 00:18:31,305 --> 00:18:33,705 for a minute and then figure out ways to make sure that their 323 00:18:33,705 --> 00:18:36,745 sandboxes are airtight because they don't wanna go to jail. And 324 00:18:36,745 --> 00:18:38,985 one of them did. The rest will figure out, okay. Let's not do 325 00:18:38,985 --> 00:18:39,465 that. 326 00:18:39,625 --> 00:18:44,070 Like, the the way to regulate this stuff is to actually have a 327 00:18:44,070 --> 00:18:46,950 regulatory scheme with penalties. It's not to sort of 328 00:18:47,270 --> 00:18:52,150 build up a a... Like a like a a church of AI that comes to 329 00:18:52,150 --> 00:18:54,790 understand it and then enlighten us all. 330 00:18:55,625 --> 00:19:00,025 Hans Buetow: Lina Khan Mhmm. Our our former Federal Trade 331 00:19:00,025 --> 00:19:03,705 Commission secretary responded to this in exactly the way that 332 00:19:03,705 --> 00:19:06,505 you're talking about. She said there are plenty of laws that 333 00:19:06,505 --> 00:19:10,630 the FTC has on the books already for this kind of situation. She 334 00:19:10,630 --> 00:19:14,070 said that law enforcers already have the authority to charge 335 00:19:14,070 --> 00:19:17,030 companies and their CEO for creating and releasing 336 00:19:17,030 --> 00:19:20,735 dangerous, unvetted, or defective products. We shouldn't 337 00:19:20,735 --> 00:19:23,455 let discussions about new legal regimes distract from the fact 338 00:19:23,455 --> 00:19:27,615 that there's no AI exemption, no exemption, from laws already on 339 00:19:27,615 --> 00:19:31,055 the books, things like consumer protection laws against unfair, 340 00:19:31,055 --> 00:19:34,175 deceptive acts, unfair methods of competition, conflict of 341 00:19:34,175 --> 00:19:38,440 interest laws, laws against poor data security, all of those 342 00:19:38,440 --> 00:19:39,720 already on the books. 343 00:19:39,800 --> 00:19:44,680 Yeah. We know, though, that we are currently being led by an 344 00:19:44,680 --> 00:19:47,935 administration that is in no way going to enforce those things. I 345 00:19:47,935 --> 00:19:51,055 think we can all kind of agree that that's just not on the 346 00:19:51,055 --> 00:19:55,775 table. Yeah. Is there then a way to read kind of what happens 347 00:19:55,775 --> 00:20:00,975 next in the timeline of the past week as the company's coming out 348 00:20:00,975 --> 00:20:02,495 and being like, okay then. 349 00:20:03,720 --> 00:20:07,800 Why don't we do it ourselves? Because on Friday of last week, 350 00:20:07,800 --> 00:20:11,720 this is just a couple of days after Cox sends his tweet, Dario 351 00:20:11,720 --> 00:20:15,080 Amade, who's the CEO of Anthropic, he comes out with a 352 00:20:15,080 --> 00:20:21,475 3,200 word blog post all about this, where he actually talks 353 00:20:21,475 --> 00:20:26,035 about the the hugging face thing and says that it's... You know, 354 00:20:26,035 --> 00:20:29,075 he gets all doomerist about it, and he says... And he agrees 355 00:20:29,075 --> 00:20:32,850 with Coxen, and he says we need a slowdown. And that he's gonna 356 00:20:32,850 --> 00:20:35,890 open his company to auditors, and that other companies should 357 00:20:35,890 --> 00:20:39,010 too, and that we're gonna pace the frontier. 358 00:20:40,050 --> 00:20:45,170 And I wanna know, like, it feels to me like this is the sort of 359 00:20:45,170 --> 00:20:48,825 thing where he's like, hey. Hey. Hey. Hey. The smart guys are 360 00:20:48,825 --> 00:20:49,385 talking. 361 00:20:49,465 --> 00:20:52,665 Don't worry. We'll... We will take care of this. That, like, 362 00:20:52,665 --> 00:20:56,185 take your cues for how all this works from us, how you feel, and 363 00:20:56,185 --> 00:20:58,665 how you behave towards our products. You should take your 364 00:20:58,665 --> 00:20:59,385 cues from us. 365 00:20:59,385 --> 00:21:01,700 And you know what? We'll just be in charge of our own regulation. 366 00:21:01,700 --> 00:21:05,220 He actually says in this thing, unfortunately, passing laws can 367 00:21:05,220 --> 00:21:08,660 take time. Mhmm. And AI is advancing very quickly. 368 00:21:08,740 --> 00:21:12,020 Therefore, in parallel with the regulatory route, AI companies 369 00:21:12,020 -->

00:21:16,865 can and should voluntarily work together to set standards. Am I 370 00:21:16,865 --> 00:21:19,345 off on that? Does it feel like they're stepping forward and 371 00:21:19,345 --> 00:21:22,945 being like, nothing... I mean, it's wrong, but we'll take care 372 00:21:22,945 --> 00:21:23,505 of it. 373 00:21:23,585 --> 00:21:28,865 Dave Karpf: So I think I have three notes. Maybe more than 374 00:21:28,865 --> 00:21:32,920 that, but let's start with three. Okay. First is I think it 375 00:21:32,920 --> 00:21:39,080 would be really interesting to give truth serum to the AI execs 376 00:21:39,320 --> 00:21:43,560 and ask them in a scenario where they had to give an honest 377 00:21:43,560 --> 00:21:47,315 answer, which are you actually more afraid of? Super 378 00:21:47,315 --> 00:21:50,035 intelligent AI or Lena Khan? 379 00:21:50,195 --> 00:21:54,195 Part of a big reason why so much of Silicon Valley moved over to 380 00:21:54,195 --> 00:21:59,475 Trump in 2024 was out of a fear of Lena Khan and other 381 00:21:59,980 --> 00:22:03,020 regulators in the Biden administration who were treating 382 00:22:03,020 --> 00:22:06,220 Silicon Valley as though it wasn't, like, the world's 383 00:22:06,220 --> 00:22:09,180 special boy who the rules didn't apply to. Yes. So, like, Lena 384 00:22:09,180 --> 00:22:11,580 Khan spent several years... There... There's this one memo. 385 00:22:11,580 --> 00:22:14,295 This, by the way, is is in my book, Vaporware Incorporated. 386 00:22:14,295 --> 00:22:15,015 Hans Buetow: I love it. I love 387 00:22:15,015 --> 00:22:18,055 Dave Karpf: it. Much better. But one of the things that I that I 388 00:22:18,055 --> 00:22:21,255 point out in that, there was a memo that got published by 389 00:22:21,255 --> 00:22:25,415 Khan's FTC that just said to the AI companies, like, hey. Just a 390 00:22:25,415 --> 00:22:28,320 reminder, you're not allowed to lie about what your products do. 391 00:22:28,320 --> 00:22:31,440 If you say that your product uses AI for some purpose, it 392 00:22:31,440 --> 00:22:33,200 actually has to use AI for that purpose. 393 00:22:33,440 --> 00:22:36,720 Otherwise, we will investigate and find you. And all of Silicon 394 00:22:36,720 --> 00:22:40,045 Valley lost their minds. Yeah. Because, like, we're supposed to 395 00:22:40,045 --> 00:22:42,045 move fast and break things. We're supposed to, you know, do 396 00:22:42,045 --> 00:22:42,605 vaporware. 397 00:22:42,605 --> 00:22:44,765 We're supposed to build up products and then wait to 398 00:22:44,765 --> 00:22:46,765 actually create the real product until after we've seen that 399 00:22:46,765 --> 00:22:49,245 there's a market for it. Like, that's our way. We're we're the 400 00:22:49,245 --> 00:22:51,645 world's special boys building the future. And they were 401 00:22:51,645 --> 00:22:55,630 outraged at this idea that the largest companies on Earth would 402 00:22:55,630 --> 00:22:58,750 be treated on the basis of their results rather than

their 403 00:22:58,750 --> 00:23:02,190 ambitions. Like, that's part of how we got here is so many of 404 00:23:02,190 --> 00:23:04,830 the AI figures, not not Amodai, but an awful lot, you know, like 405 00:23:04,830 --> 00:23:09,245 Elon Musk, you know, Greg Brockman from OpenAI, like, they 406 00:23:09,245 --> 00:23:12,925 all have donated heavily to Trump and the regime. 407 00:23:12,925 --> 00:23:16,045 A bunch of them got into Doge and led Doge because they 408 00:23:16,045 --> 00:23:18,605 believe that the regulatory state should be decimated and 409 00:23:18,605 --> 00:23:21,800 Silicon Valley should be left to its own devices. Mhmm. So while 410 00:23:21,800 --> 00:23:24,440 I do think that they believe that they're building super 411 00:23:24,440 --> 00:23:27,880 intelligent AI and maybe it'll destroy the world, I I honestly 412 00:23:27,880 --> 00:23:31,080 think their dreams are a lot more haunted by Lena Khan and 413 00:23:31,080 --> 00:23:33,800 competent regulators. It's like that's the first thing I wanna 414 00:23:33,800 --> 00:23:34,600 point out here. 415 00:23:34,600 --> 00:23:35,320 Hans Buetow: Yeah. Okay. 416 00:23:35,615 --> 00:23:38,015 Dave Karpf: That gets to the second that... I I read that 417 00:23:38,015 --> 00:23:41,855 Amadai essay. Yeah. I thought... I I saw it on Saturday, which 418 00:23:41,855 --> 00:23:43,855 meant that I had to spend my weekend reading that thing. 419 00:23:43,855 --> 00:23:50,270 Thanks a lot, buddy. But... I I read that. I've got... Wrote a 420 00:23:50,270 --> 00:23:51,630 thread responding to it on Blue Sky. 421 00:23:51,630 --> 00:23:55,230 That thread is gonna turn into an essay on tech policy press, I 422 00:23:55,230 --> 00:23:57,630 think, later on this week. But the the main thing that stands 423 00:23:57,630 --> 00:24:00,910 out, like, you... You're right. It's accurate that his stance 424 00:24:00,910 --> 00:24:04,795 is, well, the government's not gonna do it, so let's just trust 425 00:24:04,795 --> 00:24:08,555 the AI industry to regulate itself. His ask from government 426 00:24:08,555 --> 00:24:12,955 is he thinks for the big AI firms to freeze things in place 427 00:24:12,955 --> 00:24:16,075 and slow things down and handle it all, what they need is a 428 00:24:16,075 --> 00:24:19,150 waiver from the government not to pursue them for antitrust 429 00:24:19,150 --> 00:24:20,110 violations. 430 00:24:20,110 --> 00:24:22,910 So the one thing they want from regulators is really, really 431 00:24:22,910 --> 00:24:26,510 promise not to regulate us at all. Mhmm. He's... So like, 432 00:24:26,510 --> 00:24:28,910 that... Then like that stands out because again, it's it's a 433 00:24:28,910 --> 00:24:32,990 weird thing when Amadai who kind of plays the role of at least if 434 00:24:32,990 --> 00:24:35,995 not the good AI guy, then at least the less bad AI guy. 435 00:24:36,235 --> 00:24:39,755 Right? Like, when he comes forward and says, yes. You know, 436 00:24:39,755 --> 00:24:43,995 we need to slow down the pace of AI adoption. And he he has a 437 00:24:43,995 --> 00:24:47,195 line in there that, like, unfettered competition amongst 438 00:24:47,195 --> 00:24:50,740 companies is gonna be a race to the bottom. Yes. 439 00:24:50,740 --> 00:24:52,980 Like, I read that, and I think, yes. Like, this is the reason 440 00:24:52,980 --> 00:24:56,020 why I think we ought to keep on slowing down the pace of data 441 00:24:56,020 --> 00:25:00,420 centers because the the way this stuff is getting developed is 442 00:25:00,420 --> 00:25:03,780 nuts and is bad for society. So I... Like, I want to agree with 443 00:25:03,780 --> 00:25:08,515 him, And then the place where he definitely loses me is I look at 444 00:25:08,515 --> 00:25:12,195 the trajectory of this industry and think we cannot trust the AI 445 00:25:12,195 --> 00:25:16,115 industry to police the excesses and the worst behaviors of the 446 00:25:16,115 --> 00:25:18,850 AI industry. There's too much money involved. 447 00:25:18,850 --> 00:25:21,810 If if we build out this technology just through the 448 00:25:21,810 --> 00:25:25,650 mechanisms of market competition in unfettered capitalism, that's 449 00:25:25,650 --> 00:25:27,890 the way to end up at an awful lot of the worst results. Right? 450 00:25:27,890 --> 00:25:30,945 Like, that's how we end up just figuring out what is the market 451 00:25:30,945 --> 00:25:33,425 for AI murder drones, whether they work or not, because 452 00:25:33,425 --> 00:25:36,305 there's an awful lot of money in military tech. Rather than 453 00:25:36,305 --> 00:25:38,705 figuring out, okay, what's the transparency requirements? What 454 00:25:38,705 --> 00:25:39,905 are the boundaries gonna be? 455 00:25:39,985 --> 00:25:42,945 Like, who is liable when the AI murder drone kills people it 456 00:25:42,945 --> 00:25:45,420 wasn't supposed to kill? Like, all that needs to get worked 457 00:25:45,420 --> 00:25:50,220 out. The industry is not gonna do that to itself. I think that 458 00:25:50,220 --> 00:25:53,340 we as a society should be looking at AI and the pace of AI 459 00:25:53,340 --> 00:25:56,220 adoption and sort of this industry and its behaviors and 460 00:25:56,220 --> 00:25:59,645 saying, you know, you guys can't be in charge of this. We need 461 00:25:59,645 --> 00:26:00,205 governments. 462 00:26:00,205 --> 00:26:03,805 We need democracy. We need regulators to kinda get in your 463 00:26:03,805 --> 00:26:07,725 way and interfere. And he's throwing out a bunch of 464 00:26:07,725 --> 00:26:10,365 examples. Like, he's he's got this... His first proposal is 465 00:26:10,365 --> 00:26:13,960 they should have monitoring and transparency, And then he brags 466 00:26:13,960 --> 00:26:17,000 that this company, Meter, has been doing the monitoring for 467 00:26:17,000 --> 00:26:17,400 them. 468 00:26:17,720 --> 00:26:20,760 Meter is is tiny, and it's made up of a bunch of people who, 469 00:26:20,760 --> 00:26:22,840 like, live in

the same group houses as as Anthropic 470 00:26:22,840 --> 00:26:26,360 employees. Mhmm. Like, it's well meaning, but it is nowhere near 471 00:26:26,360 --> 00:26:29,365 large and powerful enough to actually be a meaningful check 472 00:26:29,365 --> 00:26:31,765 on Anthropic. And part of how you know that is if they ever 473 00:26:31,765 --> 00:26:34,245 found anything that Anthropic didn't want, Anthropic could 474 00:26:34,245 --> 00:26:37,045 could just change its mind since it's voluntary. Right. 475 00:26:37,045 --> 00:26:40,245 But he compares that to, like, the monitors in the banking 476 00:26:40,245 --> 00:26:43,880 industry. And the thing about the the monitoring in the 477 00:26:43,880 --> 00:26:47,080 banking industry is that largely comes about after the great 478 00:26:47,080 --> 00:26:48,200 financial crisis. 479 00:26:48,200 --> 00:26:48,600 Hans Buetow: Yeah. 480 00:26:48,600 --> 00:26:50,280 Dave Karpf: And then the... We have the force of government 481 00:26:50,280 --> 00:26:53,240 saying, you're gonna have this monitoring and transparency, or 482 00:26:53,240 --> 00:26:55,160 we will find the hell out of you, and there will be jail 483 00:26:55,160 --> 00:26:57,365 time. Right? Like, the force of government is the only thing 484 00:26:57,365 --> 00:27:00,965 that makes that stick. The thing he's describing is actually an 485 00:27:00,965 --> 00:27:04,645 awful lot more like Arthur Andersen and the Enron scale... 486 00:27:04,645 --> 00:27:05,685 In the Enron scandal. 487 00:27:05,685 --> 00:27:07,285 Hans Buetow: Oh, interesting comparison. Yeah. 488 00:27:07,285 --> 00:27:09,685 Dave Karpf: There... There's an awful lot of transparency in the 489 00:27:09,685 --> 00:27:13,470 Enron era. But since it's all voluntary with a third party 490 00:27:13,470 --> 00:27:16,670 that's much smaller, Enron figures out like, oh, we can 491 00:27:16,670 --> 00:27:19,470 just... It's just gonna be in the financial interest of Arthur 492 00:27:19,470 --> 00:27:21,630 Andersen to say what we want them to say. Yeah. 493 00:27:21,630 --> 00:27:24,190 That ends up furthering all of the problems. Like, you really 494 00:27:24,190 --> 00:27:27,705 do need government to make the industry feel uncomfortable. 495 00:27:28,345 --> 00:27:30,905 Otherwise, you're not gonna have any meaningful checks on the 496 00:27:30,905 --> 00:27:34,665 industry. And so, again, like, think Amadai is is well meaning. 497 00:27:34,665 --> 00:27:34,905 He... 498 00:27:34,905 --> 00:27:38,105 He's like, he wants to have morals, and those morals are in 499 00:27:38,105 --> 00:27:40,940 tension with his business ambitions, and he's just kinda 500 00:27:40,940 --> 00:27:45,020 hoping that we can make them both work. But the answer, if 501 00:27:45,020 --> 00:27:48,140 the history from any other large in... Powerful industry is any 502 00:27:48,140 --> 00:27:51,100 guide, the answer is actually you need regulations that aren't 503 00:27:51,100 --> 00:27:55,065 set by the industry itself. You need rules that they don't like 504 00:27:55,065 --> 00:27:57,305 and make them feel uncomfortable. You need serious 505 00:27:57,305 --> 00:28:00,185 liability that they can face massive fines for in jail time 506 00:28:00,185 --> 00:28:00,825 if they don't think... 507 00:28:00,825 --> 00:28:03,225 Do things wrong. Of course, the industry isn't gonna propose 508 00:28:03,225 --> 00:28:05,705 that, but that's why I think we should take his recommendation 509 00:28:05,705 --> 00:28:09,110 that we should slow the frontier and say, that's great. You don't 510 00:28:09,110 --> 00:28:11,590 get to slow the frontier. Other people do. That's right. 511 00:28:11,670 --> 00:28:13,910 But I have one more point, and this is the really bad one. 512 00:28:13,910 --> 00:28:14,710 Hans Buetow: Oh, boy. 513 00:28:15,110 --> 00:28:18,230 Dave Karpf: So the the bad one, the trap we're in Yeah. Is... 514 00:28:18,230 --> 00:28:22,365 And and you noted this. Like, we we don't have Lina Khan in the 515 00:28:22,365 --> 00:28:24,605 Kamala Harris administration. We're Yeah. 516 00:28:24,605 --> 00:28:27,405 Stuck with the Trump administration. Trump listened 517 00:28:27,405 --> 00:28:30,525 to... You know, Ahmad I published this, and we had both 518 00:28:30,525 --> 00:28:33,800 Elon Musk and Sam Altman saying, yep. We'll sign on to that. And 519 00:28:33,800 --> 00:28:35,800 then Trump replied to that by saying, no. 520 00:28:35,800 --> 00:28:38,280 We don't need any of that. All we need is a really smart 521 00:28:38,280 --> 00:28:41,320 president. We have one. So all gas, no brakes. Let's just move 522 00:28:41,320 --> 00:28:44,680 forward with it, which I think Jensen Huang seemed to really 523 00:28:44,680 --> 00:28:47,880 like because, I don't know, I guess NVIDIA needs more 524 00:28:47,880 --> 00:28:48,680 purchases or whatever. 525 00:28:49,345 --> 00:28:52,625 But we are in a situation where the Trump administration has 526 00:28:52,625 --> 00:28:55,345 gutted the entire administrative state in The United States. We 527 00:28:55,345 --> 00:28:57,025 do not have administrative capacity. We don't have 528 00:28:57,025 --> 00:29:00,465 competent regulators. They've been threatening Europe and 529 00:29:00,465 --> 00:29:03,790 basically saying, you know, give or take, if you want NATO, then 530 00:29:03,790 --> 00:29:06,670 you can't regulate anything in the AI industry. Like, we want 531 00:29:06,670 --> 00:29:09,390 American AI to to win, and we're gonna flex our muscles. 532 00:29:10,030 --> 00:29:13,230 So we're kinda in... We're kinda stuck with the worst possible 533 00:29:13,230 --> 00:29:16,475 president for this moment. Right. And that that does create 534 00:29:16,475 --> 00:29:20,555 a trap that is impossible to get out of in this moment. Right? 535 00:29:20,555 --> 00:29:23,755 Like, if I was advising Amadai, I'm like, okay. What do you do 536 00:29:23,755 --> 00:29:28,475 in the next six months? Like, the industry does need to

help 537 00:29:28,475 --> 00:29:31,490 out with slowing down since god knows this administration isn't 538 00:29:31,490 --> 00:29:36,290 gonna do it. And, also, the admin... The industry needs its 539 00:29:36,290 --> 00:29:39,730 power to be checked by somebody, and it's unclear who. 540 00:29:39,810 --> 00:29:41,970 This, again, is the reason why even though the industry won't 541 00:29:41,970 --> 00:29:46,765 like it, I'm a big fan of moratoria on data centers on 542 00:29:46,845 --> 00:29:49,485 making huge demands on these companies before they can build 543 00:29:49,485 --> 00:29:51,885 one to make sure that the communities they're in are made 544 00:29:51,885 --> 00:29:54,685 better by that data center. Like, slow all that down because 545 00:29:54,685 --> 00:29:59,085 that's one of the ways to change the trajectory of the AI future 546 00:29:59,380 --> 00:30:02,420 and take power away from these companies. If people are 547 00:30:02,420 --> 00:30:04,980 listening to this and they're they're worried, like, show up 548 00:30:04,980 --> 00:30:08,100 to a data center protest in your in your area and say, like, 549 00:30:08,100 --> 00:30:10,900 let's say no to that. That's one way to help slow down the 550 00:30:10,900 --> 00:30:14,015 trajectory of this stuff because eventually we need competent 551 00:30:14,015 --> 00:30:15,855 regulators again, but it's gonna take a while. 552 00:30:15,855 --> 00:30:17,695 Hans Buetow: It's gonna... It's... And and in the meantime, 553 00:30:17,695 --> 00:30:20,415 you know, Daria's proposing, like, oh, I'll just have 554 00:30:20,735 --> 00:30:22,975 auditors come in that'll have, you know 555 00:30:23,695 --> 00:30:25,055 Dave Karpf: Auditors who I'm friends with. 556 00:30:25,055 --> 00:30:26,815 Hans Buetow: Like Who I'm friends with. Exactly. I... Who 557 00:30:26,815 --> 00:30:29,510 I control. So don't you worry, folks. 558 00:30:29,590 --> 00:30:31,910 We're the smart... We're the geniuses who created the 559 00:30:31,910 --> 00:30:34,310 problem. We're the only geniuses who can solve the problem 560 00:30:34,550 --> 00:30:34,710 Dave Karpf: Yeah. 561 00:30:34,710 --> 00:30:36,710 Hans Buetow: Is what I hear a lot of in that. 562 00:30:36,870 --> 00:30:38,950 Dave Karpf: That that is absolutely what I read into that 563 00:30:38,950 --> 00:30:44,665 as well. And we should trust Dario Amadai just enough to say, 564 00:30:44,665 --> 00:30:47,145 alright. We're not gonna trust you for the solution. We need 565 00:30:47,145 --> 00:30:48,105 other people here. 566 00:30:48,185 --> 00:30:52,585 Hans Buetow: So let's then catapult ourselves forward a few 567 00:30:52,585 --> 00:30:55,030 days. It's alarming how fast conversation is. 568 00:30:55,030 --> 00:30:57,830 Dave Karpf: I'm so tired by this already. I'm so tired. 569 00:30:59,670 --> 00:31:01,750 Hans Buetow: Okay. Well, we're almost there because this gets 570 00:31:01,750 --> 00:31:05,190 us almost up to speed now. So let's catapult to the fifteenth. 571 00:31:05,190 --> 00:31:10,965 So this is Tuesday the fifteenth. And it seems like one 572 00:31:10,965 --> 00:31:14,165 of the responses to what you're saying where some sort of 573 00:31:14,165 --> 00:31:17,605 regulation is needed, some sort of regulation actually exists. 574 00:31:17,605 --> 00:31:22,485 We just need somebody to, like, actually employ that. Now we're 575 00:31:22,485 --> 00:31:28,030 getting these absolutely Frankenstein bizarre alliances 576 00:31:28,030 --> 00:31:34,910 of people getting together with bizarre high minded names 577 00:31:35,230 --> 00:31:35,390 Dave Karpf: Mhmm. 578 00:31:36,015 --> 00:31:39,855 Hans Buetow: To try to do something about what has erupted 579 00:31:39,855 --> 00:31:42,975 in the past week. So just this Tuesday, there was a quote, 580 00:31:42,975 --> 00:31:47,135 unquote... It was called a prohuman assembly Mhmm. In 581 00:31:47,135 --> 00:31:50,335 Washington. And this is the trio who got together. 582 00:31:51,150 --> 00:31:55,230 Coxen, who is the... Our our tweet guy from last week who no 583 00:31:55,230 --> 00:31:58,430 one knew about the eight days before that. 584 00:31:58,590 --> 00:32:00,350 Dave Karpf: Coxen Fifteen minutes of fame. 585 00:32:00,510 --> 00:32:01,950 Hans Buetow: Bernie Sanders. 586 00:32:02,030 --> 00:32:02,670 Dave Karpf: Mhmm. 587 00:32:02,670 --> 00:32:04,190 Hans Buetow: And who do you place with that? Well, of 588 00:32:04,190 --> 00:32:05,710 course, you go with Steve Bannon. 589 00:32:05,710 --> 00:32:07,705 Dave Karpf: Of course. Of course. Because otherwise, it's 590 00:32:07,705 --> 00:32:08,825 not sweaty enough. 591 00:32:11,065 --> 00:32:12,985 Hans Buetow: Guys are getting together and being like, we're 592 00:32:12,985 --> 00:32:18,745 gonna go with... We're doing prohuman AI. Yeah. What? What do 593 00:32:18,745 --> 00:32:19,225 Dave Karpf: you Yeah. 594 00:32:19,225 --> 00:32:20,610 Hans Buetow: Dave, help. 595 00:32:21,410 --> 00:32:27,170 Dave Karpf: I I I I really can't. I I I would love to sit 596 00:32:27,170 --> 00:32:32,370 down with Bernie and have a slow conversation that features an 597 00:32:32,370 --> 00:32:35,935 awful lot of me, like, peering over my glasses at him and 598 00:32:35,935 --> 00:32:40,255 saying, sorry, what? Like, just a real slow conversation with 599 00:32:40,255 --> 00:32:43,295 him about what the hell he thinks he's doing. The main 600 00:32:43,295 --> 00:32:47,770 thing that stands out to me is, like, that's all spectacle. I 601 00:32:47,770 --> 00:32:48,890 can understand how... 602 00:32:48,890 --> 00:32:51,850 Like, I can understand how Steve Bannon has decided this is a 603 00:32:51,850 --> 00:32:55,130 spectacle that I can play to. And so I'm gonna go get a bunch 604 00:32:55,130 --> 00:32:58,570 of attention before, we once again get the parable of the 605 00:32:58,570 --> 00:33:02,090 scorpion and the frog. Like, come on. It is 2026, and you've 606 00:33:02,090 --> 00:33:06,455 decided that

Steve Bannon is an honest ally for you here. Come 607 00:33:06,455 --> 00:33:07,015 on, guys. 608 00:33:07,015 --> 00:33:11,255 Yeah. But, like, this is an opportunity for them to 609 00:33:11,255 --> 00:33:15,095 grandstand, them to say, like, the world needs to wake up, and 610 00:33:15,095 --> 00:33:19,150 we need some things, and then do some performative proposals that 611 00:33:19,230 --> 00:33:22,510 sort of suck the air out of the room in this moment, and then 612 00:33:22,510 --> 00:33:26,510 the AI industry sort of continues unchecked. Because as 613 00:33:26,510 --> 00:33:29,310 substantial... Substantive threats go, this doesn't seem 614 00:33:29,310 --> 00:33:32,715 like a big one. Now again, I I I think I saw that one of the many 615 00:33:32,715 --> 00:33:36,155 bills that Bernie has introduced is a data center moratorium 616 00:33:36,155 --> 00:33:36,795 bill. 617 00:33:37,275 --> 00:33:39,675 I don't think it's gonna pass, but I'd be all for all for that. 618 00:33:39,675 --> 00:33:43,195 Like, sure, man. Go for it. Right. But an awful lot of these 619 00:33:43,195 --> 00:33:46,830 proposals that are making are starting from the premise that, 620 00:33:46,830 --> 00:33:49,630 yeah, the threat here is super intelligent AI. 621 00:33:49,710 --> 00:33:53,710 Yeah. And so we need the better AI, and we're gonna collaborate 622 00:33:53,710 --> 00:33:57,150 with the AI industry, but, like, really make some demands to do 623 00:33:57,150 --> 00:34:00,965 so... To be serious and answer that threat. And, again, going 624 00:34:00,965 --> 00:34:05,445 back to the first point, the AI industry really believes what 625 00:34:05,445 --> 00:34:08,245 they're talking about there. That doesn't mean that it's 626 00:34:08,245 --> 00:34:09,605 actually all that believable. 627 00:34:10,085 --> 00:34:13,125 There are a lot of other threats and harms from AI that come not 628 00:34:13,125 --> 00:34:17,100 from it being misaligned with humanity, but from it being a 629 00:34:17,100 --> 00:34:19,820 set of technologies that are getting applied to an awful lot 630 00:34:19,820 --> 00:34:23,100 of uses that they're not all all that good at... Good for yet. 631 00:34:23,660 --> 00:34:27,660 Like, the the problem is how we are deploying it and letting 632 00:34:27,660 --> 00:34:32,295 these unconstrained corporations go wherever the money is and 633 00:34:32,295 --> 00:34:34,775 just try to scoop the money out even if their products don't 634 00:34:34,775 --> 00:34:37,895 quite work right yet. So there's an awful lot of other threats 635 00:34:37,895 --> 00:34:43,090 that I think we ought to be more worried about, none of which fit 636 00:34:43,090 --> 00:34:46,530 the story that the AI industry wants to tell. And, like, it's 637 00:34:46,530 --> 00:34:50,050 kind of a really annoying distraction that now Bernie and 638 00:34:50,050 --> 00:34:54,850 Bannon have decided we are gonna be the the the torchbearers for 639 00:34:54,850 --> 00:34:58,175 fighting

AI because the the tune they're singing isn't gonna make 640 00:34:58,175 --> 00:34:59,695 any sense and isn't gonna do anything. 641 00:35:00,015 --> 00:35:05,215 Hans Buetow: Is there anything then to the threat of China? 642 00:35:05,215 --> 00:35:08,255 Because another thing people are calling for is AI diplomacy. 643 00:35:08,255 --> 00:35:12,610 Dario Amade in his long screed says that first we need the 644 00:35:12,610 --> 00:35:16,290 democrat... We need nations that are... You know, democracies to 645 00:35:16,290 --> 00:35:18,930 come together, then we need full international cooperation there. 646 00:35:18,930 --> 00:35:21,250 We're gonna bring it to the UN. We're gonna make this thing an 647 00:35:21,250 --> 00:35:24,530 international effort. China is, of course, the thing that's 648 00:35:24,755 --> 00:35:28,355 looming large in this, and it feels like there's IPOs is one 649 00:35:28,355 --> 00:35:34,435 pressure on these companies. And China developing something 650 00:35:34,435 --> 00:35:37,795 ambiguous that's super intelligent is on the other 651 00:35:37,795 --> 00:35:41,570 side, and they've got a... They can't slow down because one of 652 00:35:41,570 --> 00:35:43,090 those two things is gonna mean... 653 00:35:43,090 --> 00:35:45,810 Is gonna be a death knell for them if they don't keep moving. 654 00:35:46,450 --> 00:35:49,650 Dave Karpf: So there is a little something to China... The the 655 00:35:49,650 --> 00:35:51,970 China thing, but only a little something. And let me Okay. What 656 00:35:51,970 --> 00:35:55,305 I mean by that because it's important. The Amadai proposal 657 00:35:55,305 --> 00:35:58,985 that there ought to be bilateral talks between The US and China 658 00:35:59,145 --> 00:36:02,825 on the pace and direction of AI adoption, that is, in broad 659 00:36:02,825 --> 00:36:03,705 strokes, a good idea. 660 00:36:03,705 --> 00:36:06,025 Like, there ought to be bilateral lateral conversations. 661 00:36:06,025 --> 00:36:09,390 Yep. The reason why there should be bilateral conversations is 662 00:36:09,630 --> 00:36:12,990 China... The China threat is functioning as a boogeyman. 663 00:36:12,990 --> 00:36:13,390 Okay. 664 00:36:13,390 --> 00:36:16,510 Like, if the threat of China winning this race didn't exist, 665 00:36:16,510 --> 00:36:19,790 the AI industry would need to invent it because it is really 666 00:36:19,790 --> 00:36:24,725 useful to have a boogeyman that serves that... Look. Whenever 667 00:36:24,725 --> 00:36:27,045 somebody says, like, yeah, we should slow down the pace of 668 00:36:27,045 --> 00:36:30,805 data centers. We should apply liabilities to these companies. 669 00:36:30,805 --> 00:36:33,285 The answer is always, well, if you do that, then China's gonna 670 00:36:33,285 --> 00:36:33,525 win. 671 00:36:33,525 --> 00:36:36,270 Like, they need that boogeyman in order to make sure that the 672 00:36:36,270 --> 00:36:39,230 regulators take a light touch with them and allow them to sort 673

00:36:39,230 --> 00:36:42,910 of act now and apologize later. The... There's a historical 674 00:36:42,910 --> 00:36:46,430 parallel that always stands out to me. And, again, it's it's in 675 00:36:46,430 --> 00:36:48,270 the book. It's in chapter two of my forthcoming book. 676 00:36:48,935 --> 00:36:52,535 But the historical parallel is actually the the science of 677 00:36:52,535 --> 00:36:54,055 human cloning in the nineteen nineties. 678 00:36:54,055 --> 00:36:54,695 Hans Buetow: Okay. 679 00:36:55,015 --> 00:36:57,495 Dave Karpf: So nineteen nineties, you know, they they 680 00:36:57,495 --> 00:37:00,775 clone a sheep, the sheep dolly, and I think it was 1994. Yeah. 681 00:37:00,775 --> 00:37:06,390 And pretty quickly, a bunch of people who are sort of vaguely 682 00:37:06,390 --> 00:37:08,870 in the area of science, but not actual scientists in this area, 683 00:37:08,870 --> 00:37:11,270 say like, well, you know, if you can clone a sheep, then, like, 684 00:37:11,270 --> 00:37:13,270 you can clone a human. Like, it's the same thing. I'm sure 685 00:37:13,270 --> 00:37:14,470 we'll be able to do that, like, next year. 686 00:37:15,405 --> 00:37:18,285 And people freak out because, well, if we're cloning humans, 687 00:37:18,285 --> 00:37:20,845 then that seems like a radical change to what it means to be 688 00:37:20,845 --> 00:37:25,645 human. I remember this. And Bill Clinton then passes a moratorium 689 00:37:25,645 --> 00:37:30,010 saying there will be no use of government funds to advance 690 00:37:30,010 --> 00:37:34,170 human cloning. Like, you you can't do it. At that point, the 691 00:37:34,170 --> 00:37:36,650 sort of nascent tech industry, this showed up in Wired magazine 692 00:37:36,650 --> 00:37:38,090 a bunch, so I had to read all about it. 693 00:37:38,090 --> 00:37:41,365 But, like, the nascent tech industry freaked out because, 694 00:37:41,365 --> 00:37:43,925 again, it was government saying, no. No. There are regulations 695 00:37:43,925 --> 00:37:46,405 that apply to you. There are things that you can't do. And 696 00:37:46,405 --> 00:37:47,365 they don't like that story. 697 00:37:47,365 --> 00:37:49,765 They like the story of we are the people who build in the 698 00:37:49,765 --> 00:37:52,485 future. We are special. The rules don't apply to us. So they 699 00:37:52,485 --> 00:37:55,285 got real, real mad for several years. And the boogeyman they 700 00:37:55,285 --> 00:37:57,790 kept on invoking was, well, look. 701 00:37:57,790 --> 00:38:00,830 China doesn't have these limitations. So if you don't let 702 00:38:00,830 --> 00:38:05,470 us spend tax dollars to advance our science of human cloning, 703 00:38:05,950 --> 00:38:08,830 then China's gonna invent the science of human cloning. And, 704 00:38:08,830 --> 00:38:11,150 like, oh my god. Then then they'll rule. We can't allow 705 00:38:11,150 --> 00:38:11,310 that. 706 00:38:11,775 --> 00:38:14,575 And they said that for, like, about... Probably about eight 707 00:38:14,575 --> 00:38:18,415 years. They they kept on beating that drum. In retrospect, it's 708 00:38:18,415 --> 00:38:21,615 not that, like, China didn't have the same limitations as us. 709 00:38:21,775 --> 00:38:25,760 But China in the same years as the the one child policy was in 710 00:38:25,760 --> 00:38:29,040 fact not trying to build out a new technology that would let 711 00:38:29,040 --> 00:38:30,640 them produce tons and tons of humans. 712 00:38:30,720 --> 00:38:33,680 So, like, they did an awful lot of investigation into... Like, 713 00:38:33,680 --> 00:38:36,480 they they supported scientists, like, looking into stem cell 714 00:38:36,480 --> 00:38:39,265 research and a bunch of other research areas, but they were 715 00:38:39,265 --> 00:38:43,345 not on the trajectory that the pretend China boogeyman was 716 00:38:43,345 --> 00:38:45,345 supposedly on because the Chinese government didn't wanna 717 00:38:45,345 --> 00:38:49,505 do that. In the AI race today, like, yes, it is true that there 718 00:38:49,505 --> 00:38:53,460 are Chinese companies that are moving forward on on AI. There 719 00:38:53,460 --> 00:38:55,860 are really impressive models that are a few months behind our 720 00:38:55,860 --> 00:38:59,300 frontier models. And there is this concern from the AI 721 00:38:59,300 --> 00:39:02,340 industry that, like, if we don't win the AI race, the China 722 00:39:02,340 --> 00:39:02,980 would. 723 00:39:04,100 --> 00:39:07,815 And, also, I'm not a China expert, but from what I 724 00:39:07,815 --> 00:39:10,775 understand of the Chinese government, they may be 725 00:39:10,775 --> 00:39:13,575 interested in surveilling the populace, but they're not really 726 00:39:13,575 --> 00:39:17,575 interested in overturning all of humanity and all of government 727 00:39:17,720 --> 00:39:21,960 by unleashing, like, super intelligent AIs that can replace 728 00:39:21,960 --> 00:39:27,720 us. Like, that at least doesn't sound like a major goal for 729 00:39:27,720 --> 00:39:32,600 their government in a way that it's the goal for OpenAI and and 730 00:39:32,600 --> 00:39:35,115 Grok because apparently that's how they're gonna be able to 731 00:39:35,115 --> 00:39:37,995 scoop $30,000,000,000,000 out of out of the market the way, you 732 00:39:37,995 --> 00:39:41,755 know, the way these companies want to. So the China race is... 733 00:39:41,755 --> 00:39:45,035 Isn't good, and it would be a good idea to have bilateral 734 00:39:45,035 --> 00:39:47,835 talks. Again, if the Harris administration was here, I would 735 00:39:47,835 --> 00:39:50,900 be saying like, yes, Harris administration, like, there's an 736 00:39:50,900 --> 00:39:52,420 awful lot of spending here. 737 00:39:52,420 --> 00:39:54,580 Yep. And there's a lot of competition between the two 738 00:39:54,580 --> 00:39:57,460 countries. You should maybe come up with some rules of the road 739 00:39:57,460 --> 00:40:00,740 and some agreements. That's generally a good idea, but the

740 00:40:00,740 --> 00:40:05,685 China boogeyman really poisons the well here. And anytime I... 741 00:40:05,685 --> 00:40:08,725 Like, when I saw Dario Amadai talking about it, that for me 742 00:40:08,725 --> 00:40:12,405 was kind of a flag of, okay, he wants them to be... He wants the 743 00:40:12,405 --> 00:40:16,480 AI industry to police itself and wants us to make sure that even 744 00:40:16,480 --> 00:40:18,480 if they slow down, they don't slow down in a way that could 745 00:40:18,480 --> 00:40:21,440 allow the China boogeyman to to get in front of them. So again, 746 00:40:21,440 --> 00:40:24,400 it sort of strikes me as an attempt to avoid accountability 747 00:40:24,400 --> 00:40:26,880 in the way that he's doing it. So, like, yeah, worry about 748 00:40:26,880 --> 00:40:29,935 China, but, like, worry about China in the right way. Like 749 00:40:30,095 --> 00:40:30,415 Right. 750 00:40:30,415 --> 00:40:33,535 They're they're not trying to build the things that Elias 751 00:40:33,535 --> 00:40:36,415 Rudkowski and Dario Ahmaday say that they're trying to build. 752 00:40:36,655 --> 00:40:39,375 Hans Buetow: You have been very helpful, Dave, at helping me 753 00:40:39,375 --> 00:40:42,730 understand the parameters. I feel like I'm even talking 754 00:40:42,730 --> 00:40:46,970 slower. I feel so much calmer. I don't know that it's helped you 755 00:40:46,970 --> 00:40:49,610 at all, but I've I've been helped. But I am... 756 00:40:49,610 --> 00:40:52,250 Okay. So I feel like I have a sense of the players. I feel 757 00:40:52,250 --> 00:40:54,570 like I have a sense of the moments for the past week or two 758 00:40:54,570 --> 00:40:58,085 weeks or, you know, two months really. But I don't have a 759 00:40:58,085 --> 00:41:02,325 sense, just like you're saying, of where we go from here Yeah. 760 00:41:02,325 --> 00:41:03,525 And and what we do. 761 00:41:03,525 --> 00:41:08,165 Like, what can you say? You wrote recently about what you 762 00:41:08,165 --> 00:41:11,640 call a realistic strategy for limiting the harms of AI. Yeah. 763 00:41:11,640 --> 00:41:14,680 And I'm wondering if you can help me feel like there's people 764 00:41:14,680 --> 00:41:16,520 thinking about this in a good way. 765 00:41:16,520 --> 00:41:19,160 Dave Karpf: Yeah. So, again, it it... It's only a partial 766 00:41:19,160 --> 00:41:22,360 strategy because what we need is time. Yeah. What we need is to 767 00:41:22,360 --> 00:41:24,520 reset the power dynamics. 768 00:41:25,185 --> 00:41:29,505 I I I am not one of the people who thinks, like, AI is gonna 769 00:41:29,665 --> 00:41:32,065 disappear tomorrow and, like, there's nothing to it. Like, 770 00:41:32,065 --> 00:41:34,305 clearly, there's a bunch of things that this set of 771 00:41:34,305 --> 00:41:39,510 technologies is useful for. And I've got a line in the book, the 772 00:41:39,510 --> 00:41:43,750 trajectory of any new technology bends toward money. Right? So we 773 00:41:43,750 --> 00:41:46,630 need

to think an awful lot about what are the financial 774 00:41:46,630 --> 00:41:50,390 incentives that companies are going to chase after. 775 00:41:50,470 --> 00:41:55,305 Right? Like, Sam Olmbeck in in 2021 wrote one of the early big 776 00:41:55,305 --> 00:41:57,945 think essays on this stuff where he... It was called Moore's Law 777 00:41:57,945 --> 00:42:00,585 for Everything. And he was saying how, like, within a few 778 00:42:00,585 --> 00:42:04,185 years, AI is going to replace lawyers. It's gonna replace 779 00:42:04,185 --> 00:42:04,825 doctors. 780 00:42:04,825 --> 00:42:09,040 It's gonna replace it's gonna replace higher education. And 781 00:42:09,040 --> 00:42:11,360 again, as a guy who studies the history of the digital future, 782 00:42:11,600 --> 00:42:15,200 for thirty years, the big tech figures have been promising that 783 00:42:15,200 --> 00:42:19,360 their technologies are going to disrupt lawyers, doctors, and 784 00:42:19,360 --> 00:42:22,915 higher education, not because those three areas are so 785 00:42:22,915 --> 00:42:26,515 disruptable, but because that's where all the money is. They're 786 00:42:26,515 --> 00:42:28,835 they're always gonna move towards where is the big money 787 00:42:28,835 --> 00:42:31,235 in these markets Interesting. Yeah. If we leave it just to 788 00:42:31,235 --> 00:42:32,115 these companies. 789 00:42:32,515 --> 00:42:37,030 So my my short term solution, it's a partial solution, is what 790 00:42:37,030 --> 00:42:40,870 I call the data center bank shot. That right now, the AI 791 00:42:40,870 --> 00:42:43,750 future is getting shaped by a small number of companies that 792 00:42:43,750 --> 00:42:46,870 are pretty much entirely unchecked and also are cash 793 00:42:46,870 --> 00:42:50,230 furnaces. Right? Like, they are spending an awful lot more money 794 00:42:50,230 --> 00:42:53,465 than they make. And their ambition is, like, we're gonna 795 00:42:53,465 --> 00:42:57,065 keep moving faster and faster to get the next infusion of cash. 796 00:42:57,065 --> 00:43:00,585 Mhmm. That's not going well for society because there's no 797 00:43:00,585 --> 00:43:03,145 checks on them, and they get to just sort of go wherever they 798 00:43:03,145 --> 00:43:07,000 want even when those ideas are bad or unsustainable or going to 799 00:43:07,000 --> 00:43:10,680 wreak havoc. I think that the only solution, at least for, 800 00:43:10,680 --> 00:43:13,000 like, citizens, right, like, because we don't have a a 801 00:43:13,000 --> 00:43:16,200 government that's able to do anything right now. I think what 802 00:43:16,200 --> 00:43:20,055 we need as citizens is to pressure these data centers, try 803 00:43:20,055 --> 00:43:22,215 to slow them down, try to shut them down, shut down their 804 00:43:22,215 --> 00:43:26,455 construction, because the data centers are the one leverage 805 00:43:26,455 --> 00:43:30,615 point we have that can affect the pace of the AI build out. 806 00:43:30,775 --> 00:43:33,610 And

if you slow down the AI build out, then this huge 807 00:43:33,610 --> 00:43:36,010 speculative economy where these company... 808 00:43:36,010 --> 00:43:38,890 Like, there are, you know, you know, multi... Multibillion 809 00:43:38,890 --> 00:43:42,090 dollar companies right now that don't have a product, but 810 00:43:42,090 --> 00:43:45,050 they've got AI engineers. That only lasts for so long. 811 00:43:45,050 --> 00:43:48,135 Eventually, they actually need to show a profit. They can only 812 00:43:48,135 --> 00:43:51,575 keep that wheel spinning so long as they keep it feeling like the 813 00:43:51,575 --> 00:43:53,895 AI future is gonna arrive really fast. 814 00:43:54,295 --> 00:43:56,455 I think you pressure the data centers and try to slow them 815 00:43:56,455 --> 00:44:00,295 down in the hopes of popping the financial bubble around AI. 816 00:44:00,695 --> 00:44:04,750 Because if you can pop that financial bubble, hopefully, the 817 00:44:04,750 --> 00:44:08,030 the fallout isn't too extreme for the economy. But, like, I 818 00:44:08,030 --> 00:44:10,990 think it's Paul Krugman has said a few times that at this point, 819 00:44:10,990 --> 00:44:13,790 the the stock market is basically like three Nvidias in 820 00:44:13,790 --> 00:44:17,725 a trench coat. Right? Like, circuit of finance deals between 821 00:44:17,725 --> 00:44:21,725 all of these companies means that if you can pop that finance 822 00:44:21,725 --> 00:44:26,205 bubble, that creates an awful lot of havoc for these companies 823 00:44:26,685 --> 00:44:29,085 similar in form to the great financial crisis. 824 00:44:29,085 --> 00:44:29,565 Hans Buetow: Right. 825 00:44:29,565 --> 00:44:31,730 Dave Karpf: And it's after the great financial crisis that the 826 00:44:31,730 --> 00:44:35,650 banks finally have to accept regulation that they didn't 827 00:44:35,650 --> 00:44:37,730 want. Right? Like, they then need to do reporting that makes 828 00:44:37,730 --> 00:44:41,410 them uncomfortable because after that, they no longer had the 829 00:44:41,410 --> 00:44:44,050 power to set their own... Like like, set the agenda for how 830 00:44:44,050 --> 00:44:47,175 they're gonna get regulated. So kinda step one is to fight 831 00:44:47,175 --> 00:44:48,695 against data centers. 832 00:44:49,095 --> 00:44:52,775 Step two is to hope that that leads to a popping of the 833 00:44:52,775 --> 00:44:55,975 finance bubble. And, I wanna remind people, when you're in a 834 00:44:55,975 --> 00:44:59,110 bubble environment, the best thing you can do is pop it 835 00:44:59,110 --> 00:45:02,150 earlier rather than later. Like, I'm not rooting for anybody to 836 00:45:02,150 --> 00:45:05,910 lose their investment dollars here, but, like, this this 837 00:45:05,910 --> 00:45:10,550 bubble, like, either it leads to the unstoppable AI future where 838 00:45:10,745 --> 00:45:12,905 everything is abundant and everyone lives for forever 839 00:45:12,905 --> 00:45:16,665 because we get uploaded to the cloud, or it's a finance bubble 840 00:45:16,665 --> 00:45:20,905 and it is going to pop. So, like, it's either Valhalla or 841 00:45:20,905 --> 00:45:23,865 these things are nuts and they need to pop sooner rather than 842 00:45:23,865 --> 00:45:27,520 later. I'm betting that we're not about to become super 843 00:45:27,520 --> 00:45:29,840 intelligent AIs that all live for forever in the cloud. 844 00:45:30,000 --> 00:45:32,960 So in that case, we should be hoping that the bubble pops 845 00:45:32,960 --> 00:45:35,840 early. So you try to pop the bubble by pressuring the data 846 00:45:35,840 --> 00:45:38,000 centers. That's the one... The leverage point that we have. 847 00:45:38,605 --> 00:45:42,285 That then resets the political dynamics, the power dynamics of 848 00:45:42,285 --> 00:45:43,005 the industry. 849 00:45:43,405 --> 00:45:45,965 And then you hope probably with Europe leading the way since 850 00:45:45,965 --> 00:45:48,605 they actually have competent regulators, you then hope that 851 00:45:48,605 --> 00:45:52,150 regulators come in and say, well, okay. We're gonna need to 852 00:45:52,150 --> 00:45:54,710 set some rules and some liabilities and some boundaries. 853 00:45:55,110 --> 00:45:58,310 And no, like Dario and Sam Altman and Elon Musk, you don't 854 00:45:58,310 --> 00:46:00,710 get to pick and choose the regulations of the regulators. 855 00:46:00,950 --> 00:46:03,510 And also, sort of keep in mind that that weird information 856 00:46:03,510 --> 00:46:07,455 asymmetry where people with the most technical knowledge about 857 00:46:07,455 --> 00:46:10,655 AI, the people who work at these companies, have also been 858 00:46:10,655 --> 00:46:14,095 engaging in such intense groupthink for so long now. 859 00:46:14,095 --> 00:46:17,695 Like, they're all part of this one weird subculture that even 860 00:46:17,695 --> 00:46:19,230 though they're quite knowledgeable and they're 861 00:46:19,230 --> 00:46:21,710 genuine in their concerns, that doesn't mean that they're right. 862 00:46:21,710 --> 00:46:24,910 It it means that there needs to be some check on their power and 863 00:46:24,910 --> 00:46:26,830 authority, and they're not gonna create that themselves. 864 00:46:36,615 --> 00:46:40,215 Hans Buetow: Dave Karp, thank you so much for this help. This 865 00:46:40,215 --> 00:46:45,015 is really genuinely comforting to me, but also illuminating. 866 00:46:45,015 --> 00:46:46,760 Like, I feel like I have a much better grasp. 867 00:46:47,080 --> 00:46:48,440 Dave Karpf: Great talking to you. My pleasure. 868 00:46:48,520 --> 00:46:52,520 Hans Buetow: So where can folks find you, and, what is your new 869 00:46:52,520 --> 00:46:56,360 book called one more time? I want... I'm excited to read it. 870 00:46:56,360 --> 00:46:56,680 So I want 871 00:46:56,680 --> 00:46:59,475 Dave Karpf: everyone be excited to read it. So I'm on blue sky, 872 00:46:59,475 --> 00:47:06,195 davecarp.besocial. It's davekarpf.bsky.social. I have a 873

00:47:06,195 --> 00:47:08,515 weekly newsletter that's on Beehive now. It used to be on 874 00:47:08,515 --> 00:47:09,715 Substack and moved to Beehive. 875 00:47:09,715 --> 00:47:15,580 That's davecarp.beehiiivv.com. So you can subscribe there. And 876 00:47:15,580 --> 00:47:18,700 the book is called Vaporware Incorporated, how Silicon Valley 877 00:47:18,700 --> 00:47:22,020 sells the future to control the present. It's coming out January 878 00:47:22,020 --> 00:47:25,580 12 from Norton. You can preorder it wherever books are sold. 879 00:47:25,580 --> 00:47:27,180 Hans Buetow: And we'll put links to all that stuff in the show 880 00:47:27,180 --> 00:47:28,540 notes. Dave, thanks again. 881 00:47:28,540 --> 00:47:29,260 Dave Karpf: Thanks a lot. 882 00:47:49,210 --> 00:47:51,450 Mike Rugnetta: That is the show we have for you this week. We're 883 00:47:51,450 --> 00:47:54,490 gonna be back here in the main feed on or around Thursday, 884 00:47:54,570 --> 00:47:59,770 September 24. Never Post is a listener supported podcast. Ads 885 00:48:00,005 --> 00:48:04,245 cover the operating expenses, but it is our members that pay 886 00:48:04,245 --> 00:48:07,525 for the time and resources to write, research, book guests, 887 00:48:07,525 --> 00:48:12,085 record, edit, score, and sound design every episode. If you 888 00:48:12,085 --> 00:48:16,260 wanna become a member, I give you my solemn oath that every 889 00:48:16,260 --> 00:48:20,660 dollar you send to us will go towards PDFs and sound effects, 890 00:48:20,740 --> 00:48:25,540 and we will not, I repeat, we will not usher in an apocalypse 891 00:48:25,540 --> 00:48:29,145 of any kind, no matter what Georgia is saying in the 892 00:48:29,145 --> 00:48:29,865 newsletter. 893 00:48:29,865 --> 00:48:35,465 Neverpo.st. Become a member today. Never Post's producers 894 00:48:35,465 --> 00:48:38,105 are Audrey Evans, Georgia Hampton, and the mysterious 895 00:48:38,105 --> 00:48:41,120 doctor first name, last name. Our senior producer is Hans 896 00:48:41,120 --> 00:48:44,560 Buto. Our executive producer is Jason Oberholzer, and the show's 897 00:48:44,560 --> 00:48:46,960 host, that's me, is Mike Cognetta. 898 00:48:48,080 --> 00:48:51,200 You have to begin somewhere. The devil of your empty pocket moves 899 00:48:51,200 --> 00:48:56,405 as eschar go up the artery of a hollow arm, ending on the lip of 900 00:48:56,405 --> 00:49:00,725 your dismay, it shows, in the brillo morning of a shaving 901 00:49:00,725 --> 00:49:05,045 mirror. It's that morning always, and it is that morning 902 00:49:05,045 --> 00:49:10,580 now, and you must fight, not with fists, but with an eraser. 903 00:49:10,820 --> 00:49:14,020 The duelist awaits a ham sandwich on the dock where your 904 00:49:14,020 --> 00:49:19,860 ship comes in. Excerpt of Lonely Tylenol by Peter Gietze. 905 00:49:21,940 --> 00:49:24,482 Never Post is a production of Charts and Leisure, and it's 906 00:49:24,482 --> 00:49:26,562 distributed by Radiotopia.