WEBVTT - Ai + Earthquakes 0:00:01.690 --> 0:00:03.050 This is Monks and Merrill. 0:00:03.270 --> 0:00:08.700 Hosted by Michael Monks and Chris Merrill. This is Monks 0:00:09.180 --> 0:00:09.960 and Merrill. 0:00:10.220 --> 0:00:11.059 On demand. 0:00:12.289 --> 0:00:15.690 KFI AM 640, we're live everywhere on the iHeartRadio app. 0:00:15.730 --> 0:00:19.070 You hear Monks and Meryl all week long, 12 to 3, 0:00:19.070 --> 0:00:22.280 and then you get more Monks and Meryl on Saturday. 0:00:22.340 --> 0:00:24.440 Some of our best hits from the week and fresh, 0:00:24.520 --> 0:00:26.979 brand new content for all of you Saturday late-nighters. 0:00:27.220 --> 0:00:29.120 Follow us on the social medias, at Monks and Meryl, 0:00:29.280 --> 0:00:31.139 at Monks and Meryl. Maybe some behind the scenes today 0:00:31.180 --> 0:00:34.150 of Michael Monks giving his evaluation of Season 5, Game 0:00:34.190 --> 0:00:34.620 of Thrones. 0:00:34.690 --> 0:00:36.529 You watch Battle of the Bastards and tell me there's 0:00:36.570 --> 0:00:38.370 ever been a better episode of television. 0:00:39.490 --> 0:00:39.870 I agree. 0:00:40.270 --> 0:00:40.870 You plebe. 0:00:42.670 --> 0:00:43.070 Goobers. 0:00:44.170 --> 0:00:50.020 Last week, we heard from this former employee of Anthropic. 0:00:50.050 --> 0:00:53.540 They make Claude the AI model. And we've heard the 0:00:54.000 --> 0:00:57.340 head of Anthropic for a while say that things are 0:00:57.360 --> 0:01:00.720 going to be, could be touch and go. Like even 0:01:00.740 --> 0:01:03.950 the head of Anthropic has said, we might be making 0:01:03.990 --> 0:01:05.090 Frankenstein's monster here. 0:01:05.390 --> 0:01:06.110 Exactly. 0:01:06.550 --> 0:01:08.209 And I don't know what to believe anymore because President 0:01:08.230 --> 0:01:11.010 Trump told us to be calm. That we just needed 0:01:11.030 --> 0:01:13.920 a high IQ president and everything will be kept under control. 0:01:13.959 --> 0:01:15.620 And I thought, oh, okay, we're fine. 0:01:16.940 --> 0:01:19.100 Here's what, now this Coxon guy who was the former 0:01:19.140 --> 0:01:21.319 employee who said, hey, I could kill us all by 0:01:21.360 --> 0:01:23.100 the end of the decade. Here's what he said over 0:01:23.140 --> 0:01:24.760 the weekend as he was being interviewed about that. 0:01:24.920 --> 0:01:29.400 I think the concept of building a super intelligence is, well, 0:01:29.420 --> 0:01:33.830 it's been likened to summoning an alien species. So I 0:01:33.870 --> 0:01:37.270 think thinking about it in terms of an alien mind 0:01:37.830 --> 0:01:39.010 that we don't fully understand is. 0:01:40.010 --> 0:01:41.310 is pretty terrifying. 0:01:41.770 --> 0:01:44.910 This reminds me of Season 9 of The X-Files, Episode 4, 0:01:44.910 --> 0:01:46.920 when we had... No? No, it's fine. 0:01:46.940 --> 0:01:47.320 Go on. 0:01:47.319 --> 0:01:48.029 Let's hear the rest of it. 0:01:48.140 --> 0:01:50.720 I've never watched a single episode. Mike Daboski is our 0:01:50.780 --> 0:01:54.300 ABC News technology reporter, and Mike, you've been following this story. 0:01:54.760 --> 0:01:56.620 Are we doomed? Are we all going to die? What's 0:01:56.640 --> 0:01:57.180 happening with this? 0:01:57.840 --> 0:01:59.480 I'm sorry. I'm still reeling from the fact that you've 0:01:59.540 --> 0:02:02.110 never seen an episode of The X-Files. I haven't. 0:02:02.170 --> 0:02:04.550 No, you missed the previous segment, Mike, and he's not 0:02:04.610 --> 0:02:05.530 watched Game of Thrones. 0:02:06.660 --> 0:02:08.780 I did. I just don't reference it to everything that 0:02:08.800 --> 0:02:09.500 we do on the air. 0:02:09.540 --> 0:02:11.380 Well, there is a relationship to everything we do on 0:02:11.400 --> 0:02:13.280 the air to A Song of Ice and Fire. 0:02:13.300 --> 0:02:19.180 You're the worst. So I will say I have not 0:02:19.220 --> 0:02:24.920 gotten to season nine of The Exiles, but in any event. Yeah. 0:02:24.960 --> 0:02:26.780 Okay. All right. 0:02:26.919 --> 0:02:27.639 Elmer, stop. 0:02:28.800 --> 0:02:29.280 Unbelievable. 0:02:29.300 --> 0:02:29.560 Pretty good. 0:02:29.720 --> 0:02:30.040 Pretty good. 0:02:30.560 --> 0:02:30.800 All right. 0:02:30.900 --> 0:02:31.880 Mike, are we dead or what? 0:02:32.160 --> 0:02:35.530 Yep. So here's the thing. There are a number of 0:02:35.630 --> 0:02:38.410 claims that are flying around out there that were kind 0:02:38.430 --> 0:02:42.060 of initially started by this guy, Jacob Coxon, who you 0:02:42.180 --> 0:02:45.660 intro just there. He's that anthropic researcher who resigned last week. 0:02:45.700 --> 0:02:48.720 He also worked for OpenAI before that. He resigned because 0:02:48.740 --> 0:02:51.120 he said he's concerned that these companies know they're creating 0:02:51.160 --> 0:02:53.960 a technology that has the possibility of wiping out humanity 0:02:54.000 --> 0:02:55.980 by the end of the decade and that they're doing 0:02:56.040 --> 0:02:59.660 nothing to stop it. Now, later, anthropics alignment science lead 0:02:59.740 --> 0:03:01.940 Evan Hubinger said, actually put a number to it. He 0:03:01.980 --> 0:03:05.300 said that the chance of that happening is at 10%. 0:03:05.300 --> 0:03:08.320 Over the weekend, we heard from the CEO of Anthropic, 0:03:08.930 --> 0:03:11.590 Dario Amadei, who said that within six to 12 months, 0:03:12.050 --> 0:03:15.070 a swarm of AI agents could be capable of taking 0:03:15.150 --> 0:03:19.210 over the entire internet. So two different claims to evaluate here. 0:03:19.250 --> 0:03:21.790 The 10% figure, I think, is the one that gets 0:03:21.830 --> 0:03:25.030 the most attention because it is kind of the biggest 0:03:25.090 --> 0:03:28.020 and scariest of them all. The good news, I think, 0:03:28.180 --> 0:03:30.160 is that, I suppose it's good news, is that this 0:03:30.220 --> 0:03:33.529 is some pretty wild speculation, really. The 10% figure that 0:03:33.570 --> 0:03:36.930 gets cited is Likely what's known as a P-Doom figure. 0:03:37.090 --> 0:03:38.870 This is a number that people in the AI safety 0:03:38.910 --> 0:03:42.290 space use to predict the likelihood of AI eliminating human 0:03:42.330 --> 0:03:45.270 life on Earth. If you have a P-Doom, a probability 0:03:45.330 --> 0:03:48.030 doom of 10, you think there's a 10% chance of 0:03:48.050 --> 0:03:50.550 that happening. The thing about making predictions like this, though, 0:03:50.610 --> 0:03:54.200 is that they are unfalsifiable. You can't prove the negative. 0:03:54.220 --> 0:03:56.020 They can't be proven wrong. We're either going to be 0:03:56.040 --> 0:03:57.720 dead at the end of the decade or we won't be. 0:03:57.780 --> 0:04:00.260 We can't be like Dr. Strange in Infinity War and 0:04:00.280 --> 0:04:03.020 evaluate all the different possible timelines. It's kind of a 0:04:03.040 --> 0:04:05.060 one or the other sort of thing. What's more. 0:04:05.560 --> 0:04:08.550 Another pop culture reference. Our listeners are so mad today. 0:04:08.590 --> 0:04:09.250 Very good, Mike. 0:04:09.330 --> 0:04:11.350 Thank you. All right, continue. 0:04:12.090 --> 0:04:15.730 There is also no scientific process that these guys engaged 0:04:15.790 --> 0:04:18.779 in to arrive at that specific number. There's no methodology 0:04:18.900 --> 0:04:19.420 for P. 0:04:19.460 --> 0:04:19.739 Doom. 0:04:19.760 --> 0:04:24.020 There's no historical precedent either. It's basically just a vibes-based assessment. 0:04:24.060 --> 0:04:25.099 It's vibes. Okay. 0:04:25.240 --> 0:04:27.620 Jeffrey Hinton, who has worked at Google for a long 0:04:27.660 --> 0:04:30.300 time on AI, and is sometimes called the godfather of AI, 0:04:30.380 --> 0:04:33.800 he told CNN recently that anyone who estimates probabilities like 0:04:33.839 --> 0:04:36.100 that is really just making a wild guess. They're giving 0:04:36.140 --> 0:04:39.680 you their gut feeling. So, you know, we should not 0:04:39.779 --> 0:04:43.240 necessarily take that 10% number as gospel. We should dedicate 0:04:43.279 --> 0:04:45.700 maybe some brainpower to thinking about these sort of very 0:04:45.820 --> 0:04:49.700 out-there sci-fi futures, but there are some very real concerns 0:04:49.760 --> 0:04:53.029 maybe highlighted by Dario Amadei that are a little bit 0:04:53.070 --> 0:04:56.530 more in the present sphere and should be maybe taking 0:04:56.570 --> 0:04:57.549 up a little bit more of our attention. 0:04:57.790 --> 0:05:00.969 We're talking with Mike Dabusky from ABC News about the 0:05:01.150 --> 0:05:04.950 threat of AI and the potential it has to wipe 0:05:04.990 --> 0:05:07.990 us off the face of the earth sometime very soon. 0:05:08.540 --> 0:05:12.120 It's not often, Mike, that you see the operators of 0:05:12.279 --> 0:05:15.360 big-time companies, and the AI companies are the biggest of 0:05:15.400 --> 0:05:19.180 them all right now. You don't see these leaders come 0:05:19.240 --> 0:05:23.090 out and say, uh, this is dangerous. I mean, you 0:05:23.110 --> 0:05:26.130 even have Elon Musk saying maybe we should talk about 0:05:26.150 --> 0:05:27.969 some of this stuff. Can you talk a bit about 0:05:28.050 --> 0:05:31.180 what the leadership has to say about the threats? 0:05:32.180 --> 0:05:36.080 Yeah, of course. A rare moment of alignment between most 0:05:36.140 --> 0:05:39.039 of the major AI figures in the last few days 0:05:39.180 --> 0:05:43.760 around Dario Amadei's plan to address some of the concerns 0:05:43.820 --> 0:05:47.120 that were outlined. Again, he thinks that the more pressing concern, 0:05:47.160 --> 0:05:50.100 potentially within six to 12 months, is that this artificial 0:05:50.120 --> 0:05:55.670 intelligence technology could potentially take over the Internet, in his words. Now, 0:05:56.450 --> 0:05:58.770 worth also just kind of highlighting that term for a moment. 0:05:59.350 --> 0:06:02.390 That's a pretty vague term. We don't know specifically what 0:06:02.470 --> 0:06:06.599 Amadei is referencing that as. And taking over the Internet, 0:06:06.640 --> 0:06:09.140 taking down the Internet, is a really difficult thing to do. 0:06:09.180 --> 0:06:12.830 The Internet's distributed. Outages happen all the time. Sometimes they're 0:06:12.850 --> 0:06:16.560 very disruptive. Sometimes they're expensive. But the Internet ultimately continues 0:06:16.640 --> 0:06:19.320 to exist. It's very difficult to just pull the plug 0:06:19.520 --> 0:06:21.160 on it. So maybe a bit of a grain of 0:06:21.180 --> 0:06:23.660 salt with that. But even still, Amadeus put forward a 0:06:23.820 --> 0:06:30.290 three-step plan that Sam Altman from OpenAI, figures from Microsoft, Google, Anthropic, 0:06:30.410 --> 0:06:33.190 and others have kind of put themselves behind it. And 0:06:33.230 --> 0:06:36.349 he says he wants to embed independent evaluators inside every 0:06:36.490 --> 0:06:40.750 AI company with access to internal models, internal tools, and researchers. 0:06:41.120 --> 0:06:43.920 He wants to establish common safety standards among the major 0:06:44.000 --> 0:06:46.680 AI firms in democratic countries. This is where that idea 0:06:46.760 --> 0:06:49.840 of an AI slowdown comes in, what Dario Amede says 0:06:50.240 --> 0:06:53.440 calls pacing the frontier. He also says he wants to 0:06:53.480 --> 0:06:56.240 get all major world powers to sign on to these standards, 0:06:56.279 --> 0:07:00.469 including limited cooperation with China and Now, Amadei has said 0:07:00.510 --> 0:07:01.810 that that last part is probably going to be the 0:07:01.870 --> 0:07:04.370 hardest part, according to an interview he did with CBS 0:07:04.430 --> 0:07:07.910 over the weekend. What's more, a proposed slowdown does come 0:07:07.950 --> 0:07:11.310 with some real logistical hurdles. It only works if everyone 0:07:11.330 --> 0:07:13.780 agrees to it. Of course, if one company slows down 0:07:13.820 --> 0:07:16.900 their development, they get their lunch eaten. Now, at the moment, 0:07:16.980 --> 0:07:19.980 all these guys do seem to be at least ideologically 0:07:20.040 --> 0:07:23.000 or signaling that they are in line here. But whether 0:07:23.040 --> 0:07:25.700 that actually, you know, when the rubber hits the road, 0:07:25.720 --> 0:07:28.190 whether everyone's actually on board with this, I think remains 0:07:28.230 --> 0:07:31.610 to be seen. There are some concerns about potential antitrust 0:07:31.650 --> 0:07:34.630 violations as well. It's generally illegal for competitors to get 0:07:34.650 --> 0:07:37.290 together and agree to limit the output of their industry. 0:07:37.630 --> 0:07:39.610 Some critics have actually gone one step further and says 0:07:39.670 --> 0:07:43.160 this is basically regulatory capture. This is where companies basically 0:07:43.200 --> 0:07:47.000 set the tenor for their own regulation. Bigger picture here, guys. 0:07:47.220 --> 0:07:49.200 AI companies are burning through a lot of cash right now. 0:07:49.340 --> 0:07:52.680 No major AI firm makes money on this technology. And 0:07:52.860 --> 0:07:56.300 if everyone agrees to slow down on building products... They 0:07:56.320 --> 0:08:00.280 also slow down on burning cash, too. So important 30,000-foot view. 0:08:00.700 --> 0:08:04.280 While slowing down, AI may come with a potential human 0:08:04.320 --> 0:08:08.280 safety or cybersecurity upside. There's also a definite economic upside 0:08:08.300 --> 0:08:08.840 to these companies. 0:08:08.880 --> 0:08:11.200 That is a really good point. Mike Dabusky, our ABC 0:08:11.220 --> 0:08:15.400 News tech reporter. You'll follow him at Michael Dabusky on 0:08:15.420 --> 0:08:17.500 the social medias. Mike, thanks. Always good talking to you, man. 0:08:17.520 --> 0:08:18.050 Thank you so much. 0:08:18.170 --> 0:08:18.490 Thanks, Mike. 0:08:18.770 --> 0:08:19.110 Appreciate it. 0:08:19.110 --> 0:08:19.830 Of course, guys. Take care. 0:08:20.130 --> 0:08:20.310 All right. 0:08:20.330 --> 0:08:23.490 We've been told for years now that AI threatens our careers, 0:08:23.950 --> 0:08:27.820 but slowing it down could protect somebody's business who benefits 0:08:27.880 --> 0:08:30.900 when the biggest companies ask everybody to hit the brakes. 0:08:31.620 --> 0:08:33.940 That's next on Monks and Merrill after this update from 0:08:33.960 --> 0:08:35.280 the KFI 24-hour newsroom. 0:08:35.900 --> 0:08:41.540 You're listening to Monks and Merrill On Demand from KFI AM640. 0:08:41.559 --> 0:08:43.640 More stimulating talk. 0:08:44.120 --> 0:08:48.080 KFI AM 640 live everywhere on the iHeartRadio app. It's 0:08:48.179 --> 0:08:50.570 more Monks and Merrill. You hear us during the week 0:08:50.590 --> 0:08:52.330 and then you get a special dose of us on 0:08:52.370 --> 0:08:55.030 the weekends. Some fresh stuff, some of our best hits 0:08:55.110 --> 0:08:57.689 from the week. Thanks for being with us. KFI AM 0:08:57.690 --> 0:09:01.770 640 live everywhere on the iHeartRadio app. It's more Monks 0:09:01.910 --> 0:09:03.590 and Merrill. You hear us during the week and then 0:09:03.610 --> 0:09:05.780 you get a special dose of us on the weekends. 0:09:06.140 --> 0:09:08.420 some fresh stuff, some of our best hits from the week. 0:09:08.760 --> 0:09:09.390 Thanks for being with us. 0:09:09.640 --> 0:09:11.969 AI is going to kill us all that. According to 0:09:12.010 --> 0:09:14.590 a couple of people on the insides at Anthropic, one 0:09:14.610 --> 0:09:16.310 of the big two, I'm going to call them the 0:09:16.330 --> 0:09:18.290 big two. I know that you've got to Google and 0:09:18.309 --> 0:09:19.929 you've got meta and you've got all the, and you've 0:09:19.950 --> 0:09:22.429 got the grok, but I really think right now you're 0:09:22.510 --> 0:09:24.630 open AI and your Anthropics kind of lead the way 0:09:24.670 --> 0:09:27.480 when it comes to AI development. And Anthropic has been 0:09:27.520 --> 0:09:30.260 saying for a while, things are going to get bad. Uh, 0:09:30.480 --> 0:09:32.680 open AI and Sam Altman are now saying, yep, things 0:09:32.720 --> 0:09:36.580 could get bad. Um, Congress doesn't seem to be acting 0:09:36.620 --> 0:09:39.730 on this very quickly. I have a theory about this, too. 0:09:39.750 --> 0:09:42.170 But did you, Michael, did you hear Mark Rahner? I 0:09:42.190 --> 0:09:45.710 guess he did a big package on this, our news guy. 0:09:46.030 --> 0:09:46.230 Yeah. 0:09:47.429 --> 0:09:47.809 I heard it. 0:09:47.970 --> 0:09:48.210 You did? 0:09:48.230 --> 0:09:49.150 What did you think? 0:09:49.590 --> 0:09:50.340 I think we should play it. 0:09:50.530 --> 0:09:51.890 Oh, okay. 0:09:51.910 --> 0:09:54.270 Let me see if I can find this. 0:09:54.290 --> 0:09:55.330 Oh, he's talking with Ted Lieu. 0:09:56.530 --> 0:09:58.090 Representative. Ted Lieu. 0:09:58.150 --> 0:10:00.219 It's a very nice lead-in to the very final. work 0:10:00.280 --> 0:10:00.920 that Rahner did. 0:10:00.940 --> 0:10:01.100 Yeah. 0:10:15.550 --> 0:10:20.030 AI models have now gotten so advanced that they can 0:10:20.350 --> 0:10:24.970 cause catastrophic harm. And society has a way of dealing 0:10:25.170 --> 0:10:28.430 with this issue. They're called kill switches. So heavy machinery 0:10:28.480 --> 0:10:32.380 has kill switches. High-speed rail has kill switches. Jet skis 0:10:32.440 --> 0:10:35.940 have kill switches. And even your Apple iPhone. So we're 0:10:35.960 --> 0:10:39.179 just saying to these AI companies, hey, look, when you 0:10:39.220 --> 0:10:42.420 build an advanced AI model, make sure you can turn 0:10:42.480 --> 0:10:45.319 it off if it goes rogue. And then we give 0:10:45.360 --> 0:10:47.980 a process for government to order it to be shut 0:10:48.020 --> 0:10:49.600 down if it gets out of control. 0:10:49.760 --> 0:10:53.160 Over the weekend, President Trump dismissed concerns about AI. Liu 0:10:53.179 --> 0:10:55.750 told me the administration's let the industry run wild. 0:10:56.179 --> 0:10:59.210 And the administration made an intentional decision not to do 0:10:59.370 --> 0:11:02.929 any regulations. And that has been a mistake. 0:11:03.270 --> 0:11:04.530 It's harmed America. 0:11:04.730 --> 0:11:07.189 It's harming even the AI industry because now there's a 0:11:07.230 --> 0:11:10.000 massive public backlash. And that's why we need to put 0:11:10.059 --> 0:11:13.180 in common sense legislation like the AI kill switch bill, 0:11:13.240 --> 0:11:14.440 which is bipartisan. 0:11:14.520 --> 0:11:16.780 There's been reporting that even though AI leaders know the 0:11:16.860 --> 0:11:19.730 dangers of what they're creating, some are reluctant to stop 0:11:19.780 --> 0:11:21.950 out of fear some other country or company will do 0:11:21.970 --> 0:11:25.010 the work anyway, creating what I called an insane arms race. 0:11:25.320 --> 0:11:28.420 It is an insane arms race. And the example you 0:11:28.480 --> 0:11:34.940 raised with anthropic stopping biological weapons information from being given 0:11:35.600 --> 0:11:39.120 to potentially a rogue nation or terrorists, their argument is, well, 0:11:39.280 --> 0:11:41.900 if we don't do this, then China might deal with 0:11:41.960 --> 0:11:43.540 a model that can build a biological weapon. 0:11:43.559 --> 0:11:44.099 But you know what? 0:11:44.300 --> 0:11:47.360 Both are very, very bad. And so just because another 0:11:47.380 --> 0:11:51.070 country can build a virus that can kill humanity doesn't 0:11:51.090 --> 0:11:53.020 mean we should let anthropic do that, too. 0:11:53.300 --> 0:11:55.530 Lou said ultimately, AI is a tool that can be 0:11:55.590 --> 0:11:56.610 used for good or bad. 0:11:56.630 --> 0:11:59.190 And so we need to make sure that it's being 0:11:59.270 --> 0:12:01.849 used for good purposes to help society, but then to 0:12:01.950 --> 0:12:06.050 regulate any issue regarding catastrophic harm that an AI model 0:12:06.090 --> 0:12:06.590 could pose. 0:12:06.910 --> 0:12:08.330 Mark Rauner, KFI News. 0:12:08.750 --> 0:12:11.000 Mark joins us in the studio right now. Mark, that 0:12:11.120 --> 0:12:12.020 was very depressing. 0:12:12.740 --> 0:12:14.720 And you're going to get more depressed. We're only just 0:12:14.760 --> 0:12:17.520 beginning to find out about everything that's going on right now. 0:12:17.559 --> 0:12:20.359 And things are happening really fast. I mean, even since 0:12:20.420 --> 0:12:23.780 I reported that, more things have happened. But I'll tell you, 0:12:24.000 --> 0:12:26.540 just to start off on a lighter note, you should 0:12:26.580 --> 0:12:29.900 go home and watch Colossus, the Forbin Project from 1970. 0:12:29.900 --> 0:12:32.690 This is the nerdiest episode we've ever had. 0:12:32.700 --> 0:12:33.800 I thought we couldn't get nerdier. 0:12:33.860 --> 0:12:34.180 No, no, no. 0:12:34.200 --> 0:12:35.240 This has been the nerdiest episode. 0:12:35.260 --> 0:12:36.349 Ronner comes in to bring up a 1940s. 0:12:36.380 --> 0:12:39.839 It's a 1970, and I just think it's good to 0:12:39.860 --> 0:12:41.800 put a little pop culture spin on this to put 0:12:41.820 --> 0:12:45.439 it into perspective. It's about a massive supercomputer that gains 0:12:45.540 --> 0:12:47.609 sentience and takes over the world. 0:12:47.850 --> 0:12:48.590 Yeah. 0:12:48.610 --> 0:12:51.350 And the movie ends before everybody goes extinct like we're 0:12:51.390 --> 0:12:55.910 going to happen. Anyway, since I reported the story, 40 0:12:55.910 --> 0:12:59.650 Democrats have called on Mike Johnson to reconvene Congress. At 0:12:59.690 --> 0:13:03.610 the same time, the president, the attorney general have both 0:13:03.630 --> 0:13:06.670 said they're not going to regulate AI and the industry 0:13:06.690 --> 0:13:11.699 can self-police. They've downplayed the threat. I think that if 0:13:11.760 --> 0:13:14.620 the people who are developing the technology say it's a 0:13:14.640 --> 0:13:17.210 threat to us, I'm more inclined to listen to them say, 0:13:18.450 --> 0:13:20.670 than the deniers, just like I'm more inclined to say, 0:13:20.950 --> 0:13:23.689 listen to climate scientists about what's going on, rather than 0:13:23.770 --> 0:13:26.730 climate deniers who have no credentials. This seems to be 0:13:26.750 --> 0:13:30.760 a serious issue. And also, something else that may be 0:13:30.800 --> 0:13:35.380 painfully obvious is that this gerontocracy in Congress is just 0:13:35.500 --> 0:13:39.220 not equipped to deal with this issue. These are people, 0:13:39.780 --> 0:13:43.610 you think they're going to grasp recursive self-improvement in AI 0:13:43.650 --> 0:13:47.809 and autonomous hacking? These guys probably never learned how to 0:13:47.870 --> 0:13:49.489 program the timer on a VCR. 0:13:49.610 --> 0:13:51.829 So I think that is a very salient point and 0:13:51.950 --> 0:13:53.710 one I actually plan on making today. I'm glad you 0:13:53.730 --> 0:13:55.630 brought that up because I think it is one of 0:13:55.650 --> 0:13:58.110 those things where you've got such a generational divide. You're 0:13:58.170 --> 0:14:04.229 asking somebody who has no concept how to control the computers. 0:14:04.250 --> 0:14:06.309 They don't know how to turn a computer on. Don't 0:14:06.350 --> 0:14:09.470 tell me that Michael's guy there, Mitch McConnell, is out 0:14:09.490 --> 0:14:11.990 there like, I've got to fire this bad boy up 0:14:12.050 --> 0:14:13.510 so that I can develop some new code. 0:14:14.130 --> 0:14:16.490 I'm pretty sure he's been operating on AI since at 0:14:16.530 --> 0:14:17.510 least the late 90s. 0:14:17.809 --> 0:14:21.430 But you said something there a moment ago, Mark Rahner, 0:14:21.610 --> 0:14:25.970 KFI reporter, about if the developers say it's a bad idea, 0:14:25.990 --> 0:14:28.670 then maybe we should listen to them. But I mean, 0:14:28.690 --> 0:14:31.170 I got to go back to the 1940s. There's a 0:14:31.210 --> 0:14:34.790 great film that came out. No, there was a recent 0:14:34.810 --> 0:14:36.690 film about something that happened in the 1940s, but it 0:14:36.710 --> 0:14:40.330 was Oppenheimer that had reservations about the atom bomb, even saying... 0:14:40.790 --> 0:14:43.260 I am become death. Here's a guy that says it's 0:14:43.280 --> 0:14:47.700 a bad idea, but as a result of that, one 0:14:47.740 --> 0:14:51.060 might argue that we became the global leader in the world, 0:14:51.480 --> 0:14:53.450 for better or for worse, but one might argue that 0:14:53.540 --> 0:14:57.210 that put us on the footing to be, you know, 0:14:57.250 --> 0:14:59.150 basically call the shots for the globe for the next 0:14:59.330 --> 0:14:59.930 hundred years. 0:14:59.970 --> 0:15:03.490 Yeah, I've seen comparisons between the AI development and nukes, 0:15:03.530 --> 0:15:06.790 but the difference is nukes didn't gain sentience and improve 0:15:06.840 --> 0:15:13.430 themselves and act go rogue and autonomously threaten humanity. 0:15:13.530 --> 0:15:16.470 They might be used by the AI here, is my concern. 0:15:16.550 --> 0:15:18.650 Yeah, that is one of the possibilities, and it's not 0:15:18.670 --> 0:15:21.510 a very distant one either. But there are two pieces 0:15:21.550 --> 0:15:23.110 to this that we should break down. One is what 0:15:23.150 --> 0:15:25.090 we've been talking about, is that the technology is out 0:15:25.110 --> 0:15:29.570 of control and dangerous with AI systems going rogue, acting independently, 0:15:29.650 --> 0:15:33.479 and possibly leading to human extinction sooner rather than later. 0:15:33.880 --> 0:15:35.760 The other piece of this is that this could be 0:15:35.780 --> 0:15:39.340 a desperation move by these tech bros who see the 0:15:39.460 --> 0:15:44.510 AI bubble bursting, and they don't want to lose their shorts. 0:15:44.750 --> 0:15:46.570 Slow it all down, and then we don't bleed money 0:15:46.610 --> 0:15:46.910 so fast. 0:15:46.970 --> 0:15:49.190 Yes, it's been a money furnace, and the business model 0:15:49.230 --> 0:15:51.630 doesn't account for any way to make this back. People 0:15:51.670 --> 0:15:54.470 aren't buying what they're selling. They don't want it. They 0:15:54.490 --> 0:15:57.150 can't even do an IPO. They're in such sad shape. 0:15:57.170 --> 0:16:00.270 So we don't have all the facts for either of 0:16:00.310 --> 0:16:03.890 these two components yet. I think our jaws are going 0:16:03.910 --> 0:16:04.910 to drop when we finally do. 0:16:05.450 --> 0:16:10.840 I believe it was the 1993 film Jurassic Park in 0:16:10.890 --> 0:16:18.820 which famed scientist Dr. Jeff Goldblum said, your scientists were 0:16:18.860 --> 0:16:21.720 so preoccupied with whether they could, they didn't stop to 0:16:21.740 --> 0:16:22.480 think if they should. 0:16:23.280 --> 0:16:25.300 Well, that gets us back to this arms race thing, 0:16:25.350 --> 0:16:29.850 which should terrify everybody, which is that these guys know... 0:16:31.170 --> 0:16:31.610 Thank you. 0:16:31.670 --> 0:16:33.730 Welcome to Jurassic Park. 0:16:35.910 --> 0:16:36.880 Oh, you're just going to turn it off? 0:16:36.920 --> 0:16:38.859 Isn't that a great, great score? 0:16:39.460 --> 0:16:40.900 It's not my favorite of John Williams. 0:16:40.920 --> 0:16:43.840 It's my favorite. We had this conversation about a month ago. 0:16:44.140 --> 0:16:46.140 No, I don't want to go down a rabbit hole here, 0:16:46.160 --> 0:16:49.630 but John Williams' very best is Superman. It's a good one. 0:16:49.710 --> 0:16:51.110 I mean, it's not like he has a bad one. 0:16:51.210 --> 0:16:53.890 With you on that. Hey, you talked with Ted Lieu 0:16:53.910 --> 0:16:55.960 about a kill switch. What does a kill switch look like? 0:16:56.000 --> 0:16:58.360 Because I think it's really easy for us to conceptualize, 0:16:58.500 --> 0:16:59.220 just unplug it. 0:16:59.840 --> 0:17:00.200 But it's not. 0:17:00.560 --> 0:17:03.250 It's not exactly that. So he compared it to a 0:17:03.350 --> 0:17:06.230 kill switch on a jet ski or your iPhone or 0:17:06.330 --> 0:17:09.830 the iPhone. I just don't think that's necessarily, I feel 0:17:09.850 --> 0:17:10.750 like that's reductive. 0:17:10.890 --> 0:17:12.989 He went into some detail on that with me, but 0:17:13.050 --> 0:17:14.770 we don't want to get too much into the nerd 0:17:14.850 --> 0:17:17.160 weeds here or people's eyes are going to start. 0:17:17.200 --> 0:17:18.420 Why would we start that now? 0:17:18.540 --> 0:17:21.040 Glazing over. I mean, what we do know is that 0:17:21.080 --> 0:17:23.280 AI is not improving our lives as we were promised. 0:17:23.380 --> 0:17:25.340 It's putting people out of work and it's threatening us. 0:17:26.359 --> 0:17:28.660 And when you add in the data centers to all that, 0:17:28.840 --> 0:17:32.899 it's even worse. And we don't also, because things are 0:17:32.920 --> 0:17:34.880 developing so fast, we don't have a way of knowing 0:17:34.940 --> 0:17:38.820 who to believe. Like these three tech CEOs, they come 0:17:38.940 --> 0:17:43.240 out in unison and say, please regulate us. When have 0:17:43.260 --> 0:17:45.460 you known any of these people, including Musk, to be 0:17:45.520 --> 0:17:48.460 altruistic about anything? As a journalist, we have to proceed 0:17:48.480 --> 0:17:50.790 from what we can know when there's a lot we 0:17:50.869 --> 0:17:54.409 don't know. And we've never seen Musk act for the 0:17:54.490 --> 0:17:58.470 benefit of society. I mean, Carnegie gave us libraries. What's 0:17:58.550 --> 0:17:59.220 Musk given us? 0:17:59.770 --> 0:18:01.970 Well, when we know the end is near, the clock 0:18:02.030 --> 0:18:04.389 is ticking. Yeah, like 14 of them. He did give 0:18:04.430 --> 0:18:06.950 us a lot of kids. It's all going to blow up. Ronner, 0:18:07.770 --> 0:18:08.630 what is your last meal? 0:18:09.770 --> 0:18:14.490 Oh, probably steak and a vodka martini. What about you, Merrill? 0:18:15.450 --> 0:18:16.149 I'm going to live forever. 0:18:16.430 --> 0:18:23.100 Okay. I need a big plate of Cincinnati chili. 0:18:24.630 --> 0:18:25.270 Skyline Chili? 0:18:25.290 --> 0:18:26.030 I'll do Skyline. 0:18:26.210 --> 0:18:27.090 Okay. 0:18:27.330 --> 0:18:30.359 Two cheese conies and the biggest chocolate malt. 0:18:30.619 --> 0:18:33.100 Elmer, normally we would fade the music down below the 0:18:33.140 --> 0:18:34.760 people talking, but that's cool. 0:18:34.800 --> 0:18:35.359