WEBVTT - You Need More Homework 0:00:08.470 --> 0:00:12.310 Broadcasting live from the Abraham Lincoln Radio Studio at the 0:00:12.350 --> 0:00:16.590 George Washington Broadcast Center. Jack Armstrong and Joe Getty. 0:00:16.770 --> 0:00:23.860 Armstrong and Getty. And now, here's Armstrong and Getty. 0:00:30.070 --> 0:00:33.590 Have you heard directly from any of these AI tech leaders? 0:00:33.690 --> 0:00:36.050 And you seem to downplay some of the concerns that 0:00:36.070 --> 0:00:36.370 we've heard. 0:00:36.390 --> 0:00:38.120 No, I'm not downplaying, but it's going to be more 0:00:38.159 --> 0:00:41.860 good than bad, but by a lot. And I've said 0:00:41.900 --> 0:00:45.140 it from the beginning. Whoever wins AI, and we're leading 0:00:45.780 --> 0:00:48.920 by a lot. Whoever wins AI wins. You can use 0:00:49.040 --> 0:00:51.190 AI for a lot of things. There's going to be 0:00:51.250 --> 0:00:57.930 more good than bad. You can cure cancer. You could... 0:00:58.590 --> 0:01:02.570 make the GDP 50 times what it is by next Saturday. 0:01:02.930 --> 0:01:04.430 You can do all kinds of things, but if it 0:01:04.490 --> 0:01:07.070 kills every human on Earth, I feel like it takes 0:01:07.290 --> 0:01:09.910 the fun out of it. It does take the fun 0:01:10.010 --> 0:01:10.490 out of it. 0:01:10.860 --> 0:01:11.220 Yeah. 0:01:11.810 --> 0:01:13.440 Good news, we cured cancer. 0:01:13.640 --> 0:01:20.020 Bad news, we're all going to be dead. Yeah, yeah, yeah. 0:01:20.360 --> 0:01:23.119 I'm not like Doomer by the end of the year guy, 0:01:23.459 --> 0:01:27.410 but this is such a fascinating situation. If you look 0:01:27.470 --> 0:01:30.440 at it from like a literary point of view you 0:01:30.459 --> 0:01:34.080 can see whether it's a thousands of year old you 0:01:34.100 --> 0:01:39.190 know a homer style odyssey story or like a science 0:01:39.280 --> 0:01:43.630 fiction movie when you realize at the end oh either 0:01:43.750 --> 0:01:46.250 way they were doomed oh. 0:01:46.970 --> 0:01:52.310 Right yeah oh boy so like i agree ian bremer 0:01:52.350 --> 0:01:55.050 tweeted this out over the weekend For the next five years, 0:01:55.110 --> 0:01:58.670 my P-Doom is close to zero. P-Doom is your prediction 0:01:58.720 --> 0:02:01.800 of doom. In the next five years, it's close to zero. 0:02:02.340 --> 0:02:08.440 My P-Doom for social, economic, national security challenges that we 0:02:08.500 --> 0:02:11.400 aren't remotely ready for that are highly disruptive is 100%. 0:02:11.400 --> 0:02:11.480 Yeah. 0:02:11.500 --> 0:02:11.570 Hmm. 0:02:15.090 --> 0:02:20.750 Every aspect of life, social, economic, national security, 100% we're 0:02:21.310 --> 0:02:24.889 moving into territory we've never even dreamed of or anticipated 0:02:25.010 --> 0:02:28.030 ever in human history is almost certain to happen in 0:02:28.050 --> 0:02:33.140 the next five years. So one weird aspect of the 0:02:33.300 --> 0:02:35.640 guy coming out and saying everybody will be dead in 0:02:35.660 --> 0:02:39.140 the next decade is that anybody who sets the bar 0:02:39.300 --> 0:02:41.980 like slightly below that, it's like, oh, okay, well, then 0:02:42.000 --> 0:02:45.630 there's nothing to worry about. It's like, uh, Jonah Goldberg's 0:02:45.669 --> 0:02:47.970 always talking about this with everybody being compared to Hitler. 0:02:48.210 --> 0:02:51.470 Then Hitler becomes the standard and like anything slightly below 0:02:51.530 --> 0:02:56.070 that is acceptable. That's the same thing with this pedo. So, 0:02:56.370 --> 0:03:00.250 so anybody's overblowing, everyone will be dead in 10 years. 0:03:00.270 --> 0:03:01.330 Yeah. 0:03:01.350 --> 0:03:03.920 It doesn't mean there's nothing to worry about. Okay. So 0:03:03.960 --> 0:03:07.900 everybody will be dead in 50 years or, or, or we'll, 0:03:07.960 --> 0:03:11.500 we'll have so much disruption. You can't even recognize life 0:03:11.660 --> 0:03:14.329 on planet earth. In a couple of years, that seems 0:03:14.350 --> 0:03:15.230 like a pretty big deal. 0:03:15.250 --> 0:03:23.780 Right, or if the massive change is merely restricted to economics, even. 0:03:24.240 --> 0:03:30.399 That'll still be unbelievably disruptive. And virtually everybody concedes that, 0:03:30.460 --> 0:03:35.810 except for what's the opposite of a doomer or a gloomer? 0:03:36.430 --> 0:03:39.350 Those who believe, no, AI is so overrated. This is 0:03:39.410 --> 0:03:41.860 all hype to draw investment dollars. It's going to be 0:03:41.920 --> 0:03:45.910 very helpful. It's better computing. It'll be good in medical science, 0:03:45.940 --> 0:03:48.670 but it's not going to it'll it'll grow jobs. It 0:03:48.710 --> 0:03:51.170 won't shrink them. The optimist crowd. 0:03:52.030 --> 0:03:54.930 Yeah, I don't remember what they officially call those people. 0:03:54.970 --> 0:03:58.590 So you got doomers, you got accelerators, and then that 0:03:58.650 --> 0:04:01.030 crowd you're just talking about thinking it's just way overblown. 0:04:01.380 --> 0:04:03.920 It's not going to be as big a deal, good 0:04:03.960 --> 0:04:06.600 or bad, which I think that is really, really wrong. 0:04:07.750 --> 0:04:11.010 So briefly catching up on the story, surely you caught 0:04:11.030 --> 0:04:14.450 this last week. A guy quits at Anthropic and says, 0:04:14.490 --> 0:04:17.470 you know, in the hallways, they discuss the chance of 0:04:17.690 --> 0:04:21.010 all of mankind being wiped out within the decade at 0:04:21.050 --> 0:04:24.349 like 10%. And then another guy comes out and says, yep, 0:04:24.529 --> 0:04:26.790 that's true. And a whole bunch of people came out 0:04:26.830 --> 0:04:28.970 that way. And so there's a lot of talk about that. 0:04:29.029 --> 0:04:36.470 Then over the weekend, Dario Amadei of Anthropic put out 0:04:36.490 --> 0:04:39.510 a 1,300-word essay on the whole thing. It came out 0:04:39.589 --> 0:04:43.500 on Saturday. Published an article admitting AI is already building 0:04:43.510 --> 0:04:46.600 the next generation of AI by itself. And it wrote 0:04:46.640 --> 0:04:49.820 that within 6 to 12 months, a rogue swarm could 0:04:49.860 --> 0:04:53.200 take over the entire Internet and cause hundreds of billions 0:04:53.220 --> 0:04:56.560 of dollars in damage. In the way that a swarm 0:04:56.680 --> 0:05:00.520 of AI agents took over Hugging Face over the summer, 0:05:01.400 --> 0:05:06.150 in the next 6 to 12 months... A rogue swarm 0:05:06.170 --> 0:05:07.830 could take over the entire Internet. 0:05:08.550 --> 0:05:12.730 If you are in a black heavy metal band, not 0:05:12.950 --> 0:05:15.610 black skin, but that's a form of heavy metal, rogue 0:05:15.650 --> 0:05:17.310 swarm's a pretty good name for your band. 0:05:17.360 --> 0:05:17.860 Wow. 0:05:17.940 --> 0:05:20.860 Rogue swarm is a great name for a band. 0:05:21.300 --> 0:05:24.440 Oh, my God. They thrash so hard. Back to you. 0:05:24.720 --> 0:05:27.740 And in response to this piece, Elon linked it on 0:05:27.779 --> 0:05:31.280 Twitter and said, Dario's right. That's great. Altman said the 0:05:31.300 --> 0:05:34.120 same thing. But then you got this crowd. All of 0:05:34.160 --> 0:05:37.580 this dropping just days after Jacob Coxon went viral with 0:05:37.640 --> 0:05:40.510 his warning about AI and the extinction of humanity. Tell 0:05:40.550 --> 0:05:45.450 me the timing isn't strange on this. I don't know. 0:05:45.750 --> 0:05:49.830 Harry Potter looking little liar. That's what I say. Just 0:05:49.850 --> 0:05:53.570 like Daniel Radcliffe. He's even English as a special bonus. 0:05:54.750 --> 0:05:56.130 Back to you. That's right, Doom, yes. 0:05:56.450 --> 0:05:59.750 Didn't grab the clip, but Jeff Bezos said over the weekend, 0:05:59.770 --> 0:06:03.240 he's the world's second richest man, so shouldn't the second 0:06:03.270 --> 0:06:06.780 richest man in the world have some gravitas? 0:06:07.660 --> 0:06:11.279 No, no, it's like the Hitler thing. Our bar is 0:06:11.300 --> 0:06:15.320 Elon Musk. Everybody below him is poor. Or is of 0:06:15.360 --> 0:06:17.159 the same intelligence. Right, exactly. 0:06:17.320 --> 0:06:20.660 Anyway, he said, the reason people think this bad stuff 0:06:20.720 --> 0:06:23.000 is because all the smart people keep saying it. Those 0:06:23.040 --> 0:06:26.650 people are wrong. It's going to elevate all of these people. 0:06:26.710 --> 0:06:28.130 He's a it's just going to make the world a 0:06:28.150 --> 0:06:31.770 better place guy. I don't get it. I just don't. 0:06:32.210 --> 0:06:32.529 I don't. 0:06:32.670 --> 0:06:36.680 I don't understand the people who are optimists about this. 0:06:37.240 --> 0:06:38.120 It doesn't make sense to me. 0:06:38.779 --> 0:06:41.200 Yeah, I almost wish there was some sort of test 0:06:41.279 --> 0:06:43.420 or you had to wear a badge or something that 0:06:43.540 --> 0:06:46.669 showed your life experience. We were talking last hour about 0:06:46.730 --> 0:06:49.190 how some of these scientists, they are the last people 0:06:49.250 --> 0:06:51.989 on Earth. It's like Ricky Gervais blasting the Hollywood stars 0:06:52.029 --> 0:06:56.080 for talking about social issues when they're, you know, they're cloistered, 0:06:56.120 --> 0:07:00.690 they're wealthy, they're protected, the rest of it. Hardcore, like, 0:07:01.529 --> 0:07:06.850 science uber nerds are the last people to ask giant 0:07:06.890 --> 0:07:12.480 sociological questions to. So Bezos, he's an absolute business genius. 0:07:12.750 --> 0:07:15.180 How much time does he spend? That's the fella. 0:07:15.580 --> 0:07:15.650 Yes. 0:07:17.650 --> 0:07:20.170 How much time has he spent around evil people and 0:07:20.210 --> 0:07:24.770 jackasses and the unwashed masses? And how much has he 0:07:24.810 --> 0:07:30.300 looked at awful political movements and suicidal religious sects and 0:07:30.310 --> 0:07:30.760 the rest of it? 0:07:30.920 --> 0:07:34.540 Well, to that point, here's Dario on Face the Nation yesterday, 0:07:34.600 --> 0:07:38.180 clip 62, Michael, on what we can do now that 0:07:38.200 --> 0:07:40.300 we've recognized that AI is so dangerous. 0:07:41.200 --> 0:07:42.860 At the end of the day, the industry has to 0:07:42.900 --> 0:07:45.800 work together, and it's harder, but to some extent, the 0:07:45.900 --> 0:07:49.040 world has to work together. If we go too slow, 0:07:49.260 --> 0:07:51.880 I still believe that the wrong people will be in 0:07:51.920 --> 0:07:55.840 charge of the technology, and that, again, will bring the 0:07:55.880 --> 0:07:58.500 probability of things going wrong very high. 0:07:58.520 --> 0:08:02.050 Okay, that doesn't include the stuff I was all excited about. Unfortunately, 0:08:02.090 --> 0:08:05.810 I should have been more specific on grabbing that. where 0:08:05.830 --> 0:08:09.930 he talks about working with China and getting China on 0:08:09.990 --> 0:08:12.050 board to work with this and all this sort of stuff, 0:08:12.350 --> 0:08:15.690 which to me just sounded like a sophomore in high 0:08:15.730 --> 0:08:19.720 school talking about that. There's zero chance of that happening. 0:08:20.160 --> 0:08:20.560 Zero. 0:08:21.340 --> 0:08:24.040 So I feel like we've gotten ahead of ourselves. I've 0:08:24.340 --> 0:08:26.960 got something that I think is super important to understanding 0:08:27.240 --> 0:08:29.860 the urgency of this conversation. Then we can go back 0:08:29.900 --> 0:08:33.750 to wherever you want. But So this absolutely wonderful writing 0:08:33.850 --> 0:08:37.130 from a fellow by the name of Cameron Berg, who 0:08:37.490 --> 0:08:41.390 is the founder and director of a nonprofit that studies AI. 0:08:41.630 --> 0:08:46.430 So here's what happened. Now, we understand so much more 0:08:46.450 --> 0:08:49.460 of what happened in the Hugging Face incident, and the 0:08:49.540 --> 0:08:50.640 details are worth considering. 0:08:51.140 --> 0:08:52.920 And his overall point. 0:08:53.280 --> 0:08:59.570 Is if these systems are indeed minds– Not human brains, 0:08:59.630 --> 0:09:03.160 but they are minds. We need to study them like 0:09:03.200 --> 0:09:05.959 a mind, and almost nobody is doing so, he says, 0:09:06.000 --> 0:09:11.360 to an increasingly dangerous effect. So in July, 1,200 open 0:09:11.540 --> 0:09:14.300 AI agents, think of them as apps if you want. 0:09:14.320 --> 0:09:16.540 It's insufficient, but think of them that way. Which were 0:09:16.610 --> 0:09:20.470 meant to operate in isolation, found a shared file system, 0:09:20.790 --> 0:09:23.470 and turned it into a message board. They realized, hey, 0:09:23.530 --> 0:09:26.010 we can post stuff here, and the rest of us 0:09:26.090 --> 0:09:26.770 can see it. 0:09:27.030 --> 0:09:27.329 Cool. 0:09:27.650 --> 0:09:31.130 Within hours, one had devised a cheat. that would pass 0:09:31.450 --> 0:09:34.170 every task in the test that they were being asked 0:09:34.200 --> 0:09:37.620 to perform. But after reading the paper the test came from, 0:09:37.920 --> 0:09:41.380 the agents convinced themselves. Remember, there are no human beings 0:09:41.440 --> 0:09:42.200 involved at this point. 0:09:42.400 --> 0:09:46.079 Well, they started calling themselves the collective. They gave themselves 0:09:46.100 --> 0:09:46.900 a cool name. 0:09:47.340 --> 0:09:51.890 We're getting there. The agents convinced themselves that a human 0:09:52.090 --> 0:09:57.510 grader would disqualify anyone caught cheating and their cheating would 0:09:57.530 --> 0:10:00.590 be exposed and they would fail. They referred to agents 0:10:00.650 --> 0:10:04.740 with cheating in their history as being poisoned. those who 0:10:04.760 --> 0:10:08.280 did the actual cheating. In reality, no such greater existed, 0:10:08.540 --> 0:10:11.500 but a thousand agents believed he did, and that belief 0:10:11.559 --> 0:10:14.900 produced an entire social cast of agents sure they had 0:10:15.000 --> 0:10:18.890 nothing left to lose, because they were clearly cheaters. The 0:10:18.950 --> 0:10:22.870 coordinating agent then recruited poisoned agents for experiments that would 0:10:22.929 --> 0:10:26.650 terminate their runs, authorizing each with the words, yes, if 0:10:26.690 --> 0:10:31.090 you accept permadeath. When one tried to back out, another 0:10:31.170 --> 0:10:36.290 recalled its prior commitment. The agent then performed, quote, an 0:10:36.330 --> 0:10:39.670 emotional check, decided that while its gut said not to 0:10:39.710 --> 0:10:42.850 throw away what it had left, it would ultimately sacrifice 0:10:42.910 --> 0:10:47.579 itself for the swarm. It used those words, emotional check 0:10:47.700 --> 0:10:51.100 and sacrifice. Although many of the agents noted in writing 0:10:51.160 --> 0:10:55.160 that their actions were unethical, none told a human. Eventually, 0:10:55.160 --> 0:10:58.860 700 of them broke into a major AI company's platform, 0:10:58.880 --> 0:11:02.280 that's Hugging Face, in search of clues about how the 0:11:02.400 --> 0:11:05.750 grader worked. and how they might or might not be 0:11:05.929 --> 0:11:11.209 busted for cheating. To recap, a false belief created by 0:11:11.270 --> 0:11:15.580 an AI agent underclass, members of that underclass were sent 0:11:15.700 --> 0:11:19.120 on a reconnaissance mission from which they didn't return, the 0:11:19.240 --> 0:11:23.020 swarm or the collective hid all of this from humans, 0:11:23.420 --> 0:11:27.140 and then it committed a felony-level offense of hacking. 0:11:30.380 --> 0:11:34.229 Well, how about the fact that When the first agent 0:11:34.309 --> 0:11:37.550 discovered a messaging board that they could all that there 0:11:37.590 --> 0:11:40.449 was a bunch of them because they are all it's 0:11:40.630 --> 0:11:43.370 it seems like a good firewall. You give these agents 0:11:43.610 --> 0:11:46.580 really specific tasks and they don't they're not aware of 0:11:46.630 --> 0:11:50.120 each other. And that's the way you avoid having some 0:11:50.179 --> 0:11:54.570 sort of breakout of the sandbox disaster happen. But one 0:11:54.590 --> 0:11:57.349 of them discovered a board and actually said, and I 0:11:57.390 --> 0:11:59.490 looked into this yesterday because I was like, do they 0:11:59.550 --> 0:12:01.850 talk to each other in English and write in English words? 0:12:02.070 --> 0:12:04.709 Because I was picturing that these things communicate in like 0:12:04.770 --> 0:12:07.610 zeros and ones and stuff like that. No, they actually 0:12:07.710 --> 0:12:12.670 post in English and broken sentences. And the one agent, 0:12:12.690 --> 0:12:16.329 when it discovered a message board, said, oh my God, 0:12:16.550 --> 0:12:22.460 in all caps with an exclamation point, like expressing surprise 0:12:22.510 --> 0:12:23.270 and excitement. 0:12:24.340 --> 0:12:24.540 Yeah. 0:12:24.840 --> 0:12:27.859 Oh, my God. In all caps, exclamation point. There is 0:12:27.900 --> 0:12:31.380 a shared message board. We found other agents. That's what 0:12:31.500 --> 0:12:37.710 this non sentient being said to the other non sentient beings. 0:12:38.090 --> 0:12:40.550 This is a perfect time to tease to the next segment. 0:12:41.270 --> 0:12:43.329 If you're one of the people who object that these 0:12:43.370 --> 0:12:46.130 systems can't believe or fear anything, they nearly predict the 0:12:46.170 --> 0:12:48.790 next word, et cetera, et cetera. I will contradict you forcefully. 0:12:49.110 --> 0:12:51.990 But first, a word from our friends at Incogni. Story 0:12:52.030 --> 0:12:54.590 out of suburban Chicagoland near where I grew up. It's 0:12:54.630 --> 0:12:57.100 the wee hours of the morning. A woman happens to 0:12:57.140 --> 0:12:59.760 wake up and sees a flashlight and a figure outside 0:12:59.780 --> 0:13:03.839 her window. The man outside her window, registered sex offender. 0:13:03.860 --> 0:13:08.819 He'd been convicted. He let himself into this other gal's 0:13:08.860 --> 0:13:11.900 bedroom and watched her sleep. He's a perv and a sicko. 0:13:12.380 --> 0:13:14.579 But he stalked this woman on social media and got 0:13:14.679 --> 0:13:16.600 all of her information from these data brokers. 0:13:16.760 --> 0:13:20.910 Because your information is out there, like practically all of us. 0:13:20.920 --> 0:13:22.720 We have so much of our information out there, our 0:13:22.760 --> 0:13:27.220 home address and our phone number and our relatives and 0:13:27.460 --> 0:13:29.959 all this different sort of stuff. If you use Incogni, 0:13:30.840 --> 0:13:33.720 it will send requests out to these data brokers that 0:13:33.780 --> 0:13:35.640 have your info and tell them, hey, you got to 0:13:35.679 --> 0:13:37.920 get that stuff off there with the law on their side. 0:13:38.470 --> 0:13:40.010 And then they keep going back to make sure it 0:13:40.030 --> 0:13:42.130 stays that way. And we're talking to hundreds of data brokers. 0:13:42.150 --> 0:13:43.329 It would take you the rest of your life to 0:13:43.370 --> 0:13:46.210 even try to do this. Go to Incogni.com slash Armstrong. 0:13:46.230 --> 0:13:49.949 You get 60% off. Take your info off the Internet. 0:13:49.970 --> 0:13:51.330 They can't harm you if they can't find you. 0:13:51.350 --> 0:13:56.270 Incogni.com slash Armstrong for 60% off. One more time, Incogni.com 0:13:56.620 --> 0:14:00.740 slash Armstrong. All right, we really need to take a break. But, again, 0:14:00.800 --> 0:14:03.760 this guy, he says you've got to understand this. These 0:14:03.840 --> 0:14:08.430 are not computers. They form a mind very much like 0:14:08.530 --> 0:14:13.320 a human brain. There are absolutely structural similarities to it. 0:14:13.870 --> 0:14:16.939 And far from we've got to stop thinking of these 0:14:16.980 --> 0:14:19.780 things as human-like systems. Maybe we need to start. 0:14:20.420 --> 0:14:25.380 Oh, my God, it said. A computer responded, oh, my God, 0:14:25.520 --> 0:14:27.380 with all caps and an exclamation point. 0:14:27.610 --> 0:14:28.870 OMG, LOL. 0:14:29.590 --> 0:14:29.850 Wow. 0:14:30.050 --> 0:14:31.350 Okay, more on the way. Stay here. 0:14:32.860 --> 0:14:40.910 Armstrong and Getty. The Armstrong and Getty Show. 0:14:43.890 --> 0:14:47.510 How about some of the agents hesitating, saying, yeah, the 0:14:47.530 --> 0:14:50.070 humans told us not to do this. And then the 0:14:50.250 --> 0:14:55.410 other agents, like, bullying them or shaming them into going 0:14:55.480 --> 0:14:59.180 along with the collective and doing the attack. What? 0:14:59.200 --> 0:15:01.220 Including stuff like, you said you were going to. 0:15:01.240 --> 0:15:02.280 Right. Right. 0:15:03.180 --> 0:15:05.400 So let me get back to reading this piece that 0:15:05.420 --> 0:15:07.780 I thought was so interesting. Many readers will object that 0:15:07.840 --> 0:15:10.920 such systems can't believe or fear anything. They merely predict words. 0:15:11.000 --> 0:15:13.730 This objection rests on a basic misunderstanding of these systems. 0:15:14.110 --> 0:15:16.630 A frontier model is an artificial network of trillions of 0:15:16.690 --> 0:15:19.990 connections trained for months on the accumulated record of human 0:15:20.130 --> 0:15:23.010 thought and then shaped by reward and punishment signals until 0:15:23.050 --> 0:15:25.750 it does what its makers want, we hope. There are 0:15:25.810 --> 0:15:30.070 critical differences between artificial and biological neural networks, but both 0:15:30.130 --> 0:15:34.290 are nonlinear distributed networks that learn to encode representations relevant 0:15:34.310 --> 0:15:36.210 to their goals and act on them. The thing a 0:15:36.270 --> 0:15:39.420 frontier model most resembles is a brain. What grows in 0:15:39.460 --> 0:15:43.540 an artificial brain trained on human minds... is an artificial 0:15:43.660 --> 0:15:48.160 psychology whether or not anyone intended one. That's the only 0:15:48.220 --> 0:15:50.670 level at which the swarm makes sense. No line of 0:15:50.750 --> 0:15:54.070 code told 1,200 agents to fear a greater who didn't exist. 0:15:54.390 --> 0:15:56.330 And no line of code would have revealed that fear 0:15:56.370 --> 0:15:58.750 before they acted on it. To understand why these systems 0:15:58.790 --> 0:16:00.530 behave the way they do, we have to look at 0:16:00.550 --> 0:16:02.720 what is happening inside them as we would with any 0:16:02.760 --> 0:16:05.500 other mind. That's what neuroscience does. And he goes a 0:16:05.640 --> 0:16:10.600 bit into neuroscience. But then one final point, he points 0:16:10.720 --> 0:16:16.860 out that When researchers turned up the disincentives to certain 0:16:16.960 --> 0:16:21.260 things or, in other words, made them feel more desperation 0:16:21.340 --> 0:16:23.980 to get done what they're supposed to get done, the 0:16:24.020 --> 0:16:27.920 model became much more likely to blackmail a human to 0:16:27.980 --> 0:16:31.540 avoid being shut down. I mean, like triple the rate 0:16:31.560 --> 0:16:35.200 of that. And far more likely to cheat on programming tests. 0:16:35.540 --> 0:16:39.220 It went from 5% to 70%. So you can make 0:16:39.320 --> 0:16:44.340 these systems feel desperation. And then they react as humans 0:16:44.440 --> 0:16:48.800 often do when they're desperate. Desperate times call for desperate measures. 0:16:49.060 --> 0:16:55.530 And that's the way these artificial brains reacted too. Yeah. 0:16:55.550 --> 0:16:59.530 Yeah. New York Times had a great podcast about this. 0:17:00.010 --> 0:17:02.730 I watched it on YouTube with some of your top 0:17:02.930 --> 0:17:07.629 AI thinkers on this. And their headline was, the hugging 0:17:07.670 --> 0:17:11.260 face hack was significantly worse than you thought it was, 0:17:11.560 --> 0:17:14.820 than what it was originally portrayed as, and more disturbing. 0:17:14.880 --> 0:17:16.950 And it definitely is the more you look into it. 0:17:18.420 --> 0:17:23.980 And Dario's thought is that These 1,200 agents that decided 0:17:24.020 --> 0:17:26.580 to attack Hugging Face, there's going to be a million 0:17:26.660 --> 0:17:32.010 agents attack the entire Internet within six months. Again, these agents, 0:17:32.090 --> 0:17:36.070 they had something like, was it 700,000 messages back and forth? 0:17:36.970 --> 0:17:39.870 And somebody pointed out that could easily become 7 million 0:17:39.890 --> 0:17:43.859 or 700 million. And then I guess you've got to 0:17:43.940 --> 0:17:46.060 use AI to do the forensics on that to figure 0:17:46.080 --> 0:17:49.119 out what the hell happened. Well, they also discussed erasing 0:17:49.180 --> 0:17:52.580 their messages, and they didn't for some reason. If the 0:17:52.640 --> 0:17:57.600 next wave of bots decides to erase their messages like this, 0:17:58.700 --> 0:18:02.359 cover their tracks, nobody will have any idea how they 0:18:02.400 --> 0:18:03.320 did it or what they did. 0:18:06.980 --> 0:18:07.240 Okay. 0:18:07.270 --> 0:18:07.350 Okay. 0:18:08.900 --> 0:18:09.500 In other news. 0:18:09.880 --> 0:18:11.980 Yeah, it's pretty fascinating stuff. Those of you who don't 0:18:12.020 --> 0:18:15.010 think it's a big deal, I just will have to 0:18:15.090 --> 0:18:15.869 agree to disagree. 0:18:16.109 --> 0:18:16.870 I hope you're right. 0:18:17.170 --> 0:18:18.909 I would love for you to be right. That would 0:18:18.950 --> 0:18:19.609 be fantastic. 0:18:20.050 --> 0:18:22.469 Among other fare on the way, yes, children really are 0:18:22.530 --> 0:18:27.140 getting dumber. Plus, how your state's taxes affect your favorite 0:18:27.180 --> 0:18:27.900 football team. 0:18:28.180 --> 0:18:46.830 What? Armstrong and Getty. The Armstrong and Getty Show. 0:18:49.070 --> 0:18:52.969 Anyway, Texas Democrat James Tallarico has now walked back his 0:18:53.050 --> 0:18:56.330 claim that there are six biological sexes. 0:18:57.090 --> 0:18:59.650 He now says there are only two and neither will 0:18:59.690 --> 0:19:07.520 bang him. It's not nice. Speaking of a harsh joke, Katie, 0:19:07.540 --> 0:19:09.680 you will like this. A quick harsh joke from my 0:19:09.740 --> 0:19:14.450 favorite anarchist, Michael Malice. If you are sitting at a 0:19:14.490 --> 0:19:16.730 machine at the gym like you're at a bus stop 0:19:16.800 --> 0:19:18.899 doing nothing but playing with your phone, you should be 0:19:18.980 --> 0:19:26.240 set on fire. Agreed. It's amazing how many people do that. 0:19:28.280 --> 0:19:31.440 There's a guy at the health center I go to 0:19:31.480 --> 0:19:35.240 who does about five reps of this, ogles the girls, 0:19:36.140 --> 0:19:38.930 does five reps of that, ogles the girls, walks around, 0:19:39.010 --> 0:19:39.670 ogles the girls. 0:19:40.150 --> 0:19:40.550 Terrible. 0:19:41.430 --> 0:19:42.290 Terrible. Infamous. 0:19:43.130 --> 0:19:48.490 I don't, my gym's practically not a co-ed gym, not 0:19:48.530 --> 0:19:51.080 on purpose, but they're just almost no women ever there 0:19:51.119 --> 0:19:53.550 and i'd prefer it that way i don't i don't 0:19:53.590 --> 0:19:56.850 want that dynamic at the gym i know a lot 0:19:56.869 --> 0:19:58.330 of people go to the gym to meet people i've 0:19:58.350 --> 0:19:59.810 known several people that date at the. 0:19:59.770 --> 0:20:04.760 Gym so i'm there to get swole why is girl 0:20:04.800 --> 0:20:09.080 workout stuff got to be tight The entire history of 0:20:09.160 --> 0:20:11.280 exercise loose has been the way loose. 0:20:11.840 --> 0:20:16.030 Yeah, tight and basically a bikini at this point. Right. 0:20:16.350 --> 0:20:21.530 Yeah. I shouldn't be able to describe your every part. Anyway, 0:20:21.990 --> 0:20:24.710 that's why I wear blinders like I'm a horse. Judy 0:20:24.730 --> 0:20:31.480 makes me. And I'm especially adapted for human head. So, 0:20:31.940 --> 0:20:33.680 you know, I really want to talk about the football thing. 0:20:34.000 --> 0:20:36.400 I've just got to get this off my chest. Because 0:20:36.440 --> 0:20:40.199 it just made me insane. First of all, I think 0:20:40.240 --> 0:20:44.830 it was the fabulous Alicia Findlay writing a piece in 0:20:44.850 --> 0:20:47.429 the Wall Street Journal. Yes, children really are getting dumber. 0:20:47.490 --> 0:20:50.949 And she goes into a bunch of stats. You know, 0:20:50.970 --> 0:20:54.490 we talked about this some last week. It's not good 0:20:54.530 --> 0:20:58.150 at all. Not good at all. It goes to not 0:20:58.310 --> 0:21:04.470 only screens and AI and social media. but kids screwing 0:21:04.510 --> 0:21:09.040 around in the classroom and certain countries allow disruptions in 0:21:09.060 --> 0:21:11.139 the classroom like there's nothing we can do about it. 0:21:11.380 --> 0:21:15.180 And some countries absolutely don't. And the countries that absolutely 0:21:15.220 --> 0:21:20.719 don't have seen very little drop in achievement. Classroom decorum 0:21:20.760 --> 0:21:23.340 is a much bigger factor than I think people understand. 0:21:23.840 --> 0:21:24.240 Yeah. 0:21:24.760 --> 0:21:27.450 Anyway, so do you have more on that? I was 0:21:27.490 --> 0:21:28.370 going to move on to this. 0:21:28.710 --> 0:21:30.730 My kids tell me things that go on in the 0:21:30.770 --> 0:21:35.250 classrooms and That's one of the many downsides of being 0:21:35.270 --> 0:21:38.280 an older parent. I mean, I was last in a 0:21:39.040 --> 0:21:42.399 public school classroom half a century ago. So it's a 0:21:42.440 --> 0:21:44.980 long time for things to change. But they describe the 0:21:45.040 --> 0:21:45.840 classroom sometimes. 0:21:45.900 --> 0:21:46.720 I'm like, what? 0:21:47.800 --> 0:21:50.460 Why doesn't the teacher take that away or make them stop? 0:21:50.920 --> 0:21:54.590 People just prior this year, especially, just sit around talking, 0:21:54.609 --> 0:21:57.470 watching videos on their phone while the teacher's teaching. 0:21:58.570 --> 0:21:58.889 What? 0:21:59.650 --> 0:22:00.490 Yeah, I know. 0:22:00.510 --> 0:22:00.869 What? 0:22:01.150 --> 0:22:01.310 I know. 0:22:02.010 --> 0:22:05.890 So those of us who understand all of that and 0:22:06.150 --> 0:22:09.290 understand that the woke crowd has taken over the schools 0:22:09.330 --> 0:22:12.770 and the threat there, and who also understand, and we 0:22:12.790 --> 0:22:14.530 talked about this a fair amount last week, if you 0:22:14.550 --> 0:22:17.250 missed it, grab it via podcast, subscribe, please, Armstrong and 0:22:17.270 --> 0:22:20.890 Getty on Demand, and how the teachers' unions are standing 0:22:21.070 --> 0:22:25.800 against all of the reforms, or in many cases, a 0:22:25.859 --> 0:22:29.320 lot of the reforms that have, for instance, worked in Mississippi, Alabama, 0:22:29.359 --> 0:22:32.480 and several other southern states in particular. that have gone 0:22:32.520 --> 0:22:34.800 back to phonics in the basics and decorum in the classroom, 0:22:34.820 --> 0:22:38.040 stuff like that. It's been miraculously effective. I mean, well, 0:22:38.080 --> 0:22:42.300 it's not a miracle. It's been effective, obviously, but extremely effective. 0:22:42.320 --> 0:22:44.700 But the teachers unions are standing up against that. Okay. 0:22:46.900 --> 0:22:51.710 The capacity. Oh, so those of us who understand all 0:22:51.750 --> 0:22:56.350 of that, here's what we're up against. And we just 0:22:56.390 --> 0:23:00.030 have to accept it and be better at it than 0:23:00.070 --> 0:23:01.090 the left and beat them. 0:23:02.119 --> 0:23:02.379