WEBVTT - The AI Uprising Will Be Anthropomorphized - Week in Tech 0:00:12.220 --> 0:00:15.340 Natasha Taylor Reid, I thought until this week that my 0:00:15.420 --> 0:00:19.450 stepmother was the last known survivor on MapQuest, but no longer. 0:00:20.829 --> 0:00:23.550 Yes, MapQuest, the most popular download. 0:00:23.790 --> 0:00:24.169 Yeah. 0:00:24.270 --> 0:00:26.910 This is because of Lake Ontario, right? Or Lake America? 0:00:27.130 --> 0:00:30.430 Yeah, MapQuest, unlike Google Maps and Apple, who both changed 0:00:30.480 --> 0:00:34.340 Lake Ontario to Lake America, MapQuest stood loud and proud with. 0:00:35.920 --> 0:00:38.260 Lake Ontario and got a big boost in the app 0:00:38.280 --> 0:00:39.420 charts as a result. 0:00:39.580 --> 0:00:43.180 Do all these lakes and bodies of water and whatever else, 0:00:43.540 --> 0:00:46.060 do they just go back as soon as Trump leaves office? 0:00:46.200 --> 0:00:47.000 Is that what's going to happen? 0:00:47.380 --> 0:00:48.779 That's a good question. 0:00:48.800 --> 0:00:52.690 I don't know how MapQuest has decided to call the 0:00:52.729 --> 0:00:54.990 Strait of Hormuz the Strait of Trump, or maybe it's 0:00:55.030 --> 0:00:56.310 also holding firm there. 0:00:56.330 --> 0:00:57.170 Just everything is Trump. 0:01:03.470 --> 0:01:05.960 Welcome to Tech Stuff. I'm Oz Veloshian, and this is 0:01:06.000 --> 0:01:08.420 The Week in Tech, where I'm joined by the world's 0:01:08.480 --> 0:01:11.380 most plugged-in reporters to break down what's really happening in 0:01:11.420 --> 0:01:14.950 tech right now. Today, we're joined by Reid Albregotti, tech 0:01:14.970 --> 0:01:18.970 editor of Semaphore, Taylor Lorenz of UserMag, and Natasha Tiku, 0:01:19.010 --> 0:01:21.250 tech reporter at The Washington Post. Welcome all. 0:01:21.770 --> 0:01:23.030 Hey, everyone. Good to be here. 0:01:23.170 --> 0:01:23.920 Good to be back. 0:01:24.080 --> 0:01:25.120 Yeah, it's been a little break. 0:01:25.720 --> 0:01:27.760 I missed you all on the, on the beach in Greece. 0:01:30.260 --> 0:01:32.500 We haven't had a round table for a couple of weeks. Uh, 0:01:32.620 --> 0:01:35.420 it's been an eventful couple of weeks and in general, 0:01:35.440 --> 0:01:36.520 we spend quite a lot of time on the show 0:01:36.560 --> 0:01:40.080 talking about kind of the, the new guard in tech. Um, 0:01:40.270 --> 0:01:42.929 but the old guard have recently been dominating the headlines 0:01:42.970 --> 0:01:45.310 for better or for worse. Reed, why don't you kick 0:01:45.330 --> 0:01:46.470 us off with the meta settlement? 0:01:47.250 --> 0:01:52.540 Yeah. So meta settled its big tobacco, uh, attorney general lawsuit. Um, 0:01:52.530 --> 0:01:56.890 $ 18 billion. But I think the bigger story is all 0:01:56.930 --> 0:02:00.550 these new restrictions on how teens can use the service. 0:02:00.570 --> 0:02:02.210 So they're limited to two hours a day. 0:02:02.430 --> 0:02:03.330 Not just teens. 0:02:03.650 --> 0:02:06.380 Taylor's, you know, I saw you tweeting about this, Taylor. 0:02:06.390 --> 0:02:08.139 So I know you have like major thoughts. 0:02:08.480 --> 0:02:11.679 Yeah, I've been covering this very closely. What I would 0:02:11.720 --> 0:02:14.820 say is that so Meta, this is nothing. Meta does 0:02:14.840 --> 0:02:17.200 not have to pay that full $ 18 million at all, 0:02:17.260 --> 0:02:20.079 which is a drop in the bucket for Meta. I'm sorry, billion. 0:02:20.080 --> 0:02:23.200 18 million would be like... We're truly nothing. 0:02:23.320 --> 0:02:27.960 Truly nothing. Yeah, Mark Zuckerberg's lunch. Basically, what the result is, 0:02:28.040 --> 0:02:30.230 as Reid mentioned, is there's a bunch of new restrictions 0:02:30.450 --> 0:02:35.070 on anyone's account who Meta can't verify is an adult 0:02:35.270 --> 0:02:39.190 through identity verification. So Meta is going to start harvesting 0:02:39.430 --> 0:02:43.540 significantly more data on users, all users, in order to 0:02:43.680 --> 0:02:45.020 ID verify them. 0:02:45.400 --> 0:02:46.760 And how will they do that? Will it be like... 0:02:47.470 --> 0:02:49.570 driving license? So what are the ways they're going to 0:02:49.590 --> 0:02:50.150 actually be doing it? 0:02:50.230 --> 0:02:53.570 Yeah, absolutely. So if they determine that you might be 0:02:53.710 --> 0:02:56.430 underage or that you're interacting with an underage person, you 0:02:56.450 --> 0:02:59.510 will be uploading your government identification. A lot of people 0:02:59.560 --> 0:03:02.900 have their information stored already. So their payment information is 0:03:02.919 --> 0:03:06.220 another way to do it, harvesting biometric data. But they 0:03:06.240 --> 0:03:08.919 are going to be doing ID verification for the internet. 0:03:09.040 --> 0:03:12.400 It is a huge rollback. And Anybody that cannot verify 0:03:12.560 --> 0:03:16.670 their identity within 14 days basically will be put under 0:03:16.750 --> 0:03:20.310 all of these heavy restrictions. So adult accounts... anybody that 0:03:20.470 --> 0:03:23.620 does not meet these qualifications. It's very bad because these 0:03:23.660 --> 0:03:26.360 restrictions are, you know, unable to view lots of significant 0:03:26.380 --> 0:03:28.840 amounts of content that is deemed harmful for children. We 0:03:28.900 --> 0:03:32.320 know that the government has determined in other states that 0:03:32.600 --> 0:03:35.530 content that is harmful for children includes anti-ICE content. It 0:03:35.610 --> 0:03:40.470 includes LGBTQ content. It includes reproductive justice, healthcare content. 0:03:40.550 --> 0:03:40.690 So. 0:03:41.550 --> 0:03:44.970 This is a horrible loss for free speech, and it 0:03:45.010 --> 0:03:48.360 doesn't do anything to change the business model of meta. And, 0:03:49.000 --> 0:03:52.120 you know, it's basically just a way for these attorney 0:03:52.140 --> 0:03:53.660 generals to get a headline. 0:03:54.220 --> 0:03:55.640 Reid and Tasha, do you agree with Taylor's take? 0:03:56.460 --> 0:03:59.360 I do agree. I think it is kind of a rollback. 0:03:59.440 --> 0:04:02.470 I mean, look, my kids are eight and ten. Like, 0:04:02.490 --> 0:04:05.650 they're too young to use this stuff right now. So 0:04:06.350 --> 0:04:08.410 I sort of had this fantasy in my mind since 0:04:08.430 --> 0:04:10.390 they were born that, like, by the time they are— 0:04:10.920 --> 0:04:13.770 of social media age will have figured this all out. Uh, 0:04:13.850 --> 0:04:16.849 clearly that fantasy is not going to become reality. Like 0:04:16.890 --> 0:04:18.530 this is still a problem. And like, you know, we 0:04:18.570 --> 0:04:21.350 go to Australia every year by, you know, my, my 0:04:21.550 --> 0:04:24.990 in-laws are in Sydney and they passed this ban. Um, 0:04:25.410 --> 0:04:28.180 everybody was so excited about it because it was like, 0:04:28.390 --> 0:04:30.739 you know, I think parents just, it's just so hard 0:04:30.920 --> 0:04:33.159 for them to, to deal with this, to grapple with 0:04:33.240 --> 0:04:36.660 what's happening on social media. Um, I was very skeptical 0:04:36.779 --> 0:04:39.950 as Taylor is like, any of these regulations are actually 0:04:39.990 --> 0:04:42.070 going to solve this problem. And they and they may 0:04:42.150 --> 0:04:44.870 create new problems. I think one point I would make, though, 0:04:45.350 --> 0:04:49.620 is that on the ID verification point, like somehow, because 0:04:49.660 --> 0:04:53.820 of AI, because of all this, you know, just inauthenticity online, 0:04:54.640 --> 0:04:56.280 I think we're going to have to figure out some 0:04:56.360 --> 0:05:00.430 way to do privacy preserving some sort of verification of 0:05:00.850 --> 0:05:03.170 not just your age, but that you're an actual human being. 0:05:03.790 --> 0:05:06.400 And I think this is like, an opportunity, right? Like 0:05:06.420 --> 0:05:09.940 there should be a technical solution. I think there are 0:05:09.980 --> 0:05:12.760 lots of really, there are lots of really interesting technical 0:05:12.940 --> 0:05:15.900 ideas out there. I'm not, I'm not going to endorse 0:05:16.040 --> 0:05:18.990 any one of them right now. But like, there's, I 0:05:19.000 --> 0:05:20.850 think this is a great opportunity for innovation. 0:05:21.210 --> 0:05:23.990 But I feel like I have whiplash as a tech reporter, 0:05:24.029 --> 0:05:27.190 because just less than 10 years ago, they were hauling 0:05:27.250 --> 0:05:30.030 Mark Zuckerberg in front of Congress. Remember when he, you know, 0:05:30.089 --> 0:05:33.440 Zuckerberg says in the Cambridge Analytica sort of, scandal, like Senator, 0:05:33.480 --> 0:05:35.600 we sell ads, you know, and they were like, why, 0:05:35.660 --> 0:05:38.240 you know, you harvested all this data on users. And 0:05:38.820 --> 0:05:42.420 now we are mandating that they do significantly more than that. 0:05:43.350 --> 0:05:46.070 I think the issue of sort of human stuff online 0:05:46.089 --> 0:05:49.200 at like sort of determining humans versus bots, like I 0:05:49.240 --> 0:05:51.860 agree with Reid. I think that is an increasing problem 0:05:51.900 --> 0:05:53.820 and we are going to have to find some solution 0:05:53.839 --> 0:05:57.380 to that. Meta does not seem interested in that solution. 0:05:57.430 --> 0:06:00.890 In fact, we had Adam Masseri just recently talking on 0:06:00.910 --> 0:06:03.630 a podcast about how they want more AI creators and 0:06:03.650 --> 0:06:05.049 they're leaning further into AI. 0:06:05.310 --> 0:06:07.910 Taylor, there's another overarching theme. I bet you'll agree with 0:06:07.950 --> 0:06:12.010 me on this, but this applies to this FTC settlement 0:06:12.029 --> 0:06:14.830 with Amazon. It applies to the AI regulation that's being 0:06:14.890 --> 0:06:19.669 passed today. I think tech regulation needs to happen much faster. 0:06:19.970 --> 0:06:22.590 And it needs to be able to change much faster, 0:06:22.670 --> 0:06:26.030 because none of this stuff like the AI regulations that 0:06:26.050 --> 0:06:28.870 have been passed are already, like they seem like years 0:06:29.089 --> 0:06:31.920 out of date. And I think this, this one will 0:06:31.950 --> 0:06:34.680 be another one where, like, yeah, okay, let's try to, 0:06:34.880 --> 0:06:37.159 let's try to work this out. But let's, let's be nimble. 0:06:37.260 --> 0:06:39.300 Let's be able to pass laws right now. We can't 0:06:39.320 --> 0:06:41.680 do that. I mean, we can't, especially on a federal level. 0:06:42.310 --> 0:06:44.599 I think another thing is like, what are you passing 0:06:44.640 --> 0:06:48.780 laws based on? And this whole sort of child safety 0:06:48.940 --> 0:06:52.280 movement lately is passed on nothing. It's nothing but vibes. 0:06:52.380 --> 0:06:55.880 We have every single top researcher who studies the use 0:06:55.940 --> 0:06:58.860 of social media technology saying, hey, there's not a causal relationship. 0:06:59.200 --> 0:07:02.980 In fact, depressed kids use Instagram more. It's a symptom 0:07:03.180 --> 0:07:05.560 of the broader problem. We know what's causing this use 0:07:05.580 --> 0:07:08.500 mental health crisis. Let's address that stuff. And it's sort 0:07:08.520 --> 0:07:11.130 of this like after the fact, again, just like people 0:07:11.150 --> 0:07:12.970 want to put a bandaid. They want a headline. They 0:07:12.990 --> 0:07:14.910 want to get reelection. Like, look at I'm tough on 0:07:14.930 --> 0:07:18.070 big tech while you've done nothing to meaningfully change the 0:07:18.090 --> 0:07:20.070 tech landscape, but make it worse for all of us. 0:07:20.790 --> 0:07:22.860 And I agree. I mean, at this point, it's so 0:07:22.900 --> 0:07:25.600 late anyway, because these kids are busy talking to chat 0:07:25.640 --> 0:07:28.480 bots like Instagram and Snapchat is the least of of 0:07:28.520 --> 0:07:30.500 what parents should be sort of concerned about. 0:07:30.960 --> 0:07:31.180 Yeah. 0:07:31.380 --> 0:07:33.380 Natasha, I'm curious, what's what's your take on this on 0:07:33.400 --> 0:07:34.120 this meta settlement? 0:07:34.770 --> 0:07:37.830 I don't know. I remember getting excited about GDPR and 0:07:37.890 --> 0:07:42.650 feeling like, you know, maybe this will be like, this 0:07:42.690 --> 0:07:43.230 will be it. 0:07:43.530 --> 0:07:44.090 Not a lot. 0:07:44.210 --> 0:07:45.550 GDPR is the bane of my career. 0:07:45.790 --> 0:07:46.410 It's terrible. 0:07:46.770 --> 0:07:49.770 And now I just like click, like reject, accept, whatever. 0:07:50.230 --> 0:07:54.570 So it just feels not that everyone is fixated on 0:07:55.380 --> 0:07:58.200 AI solely, but it does feel like while the ball 0:07:58.280 --> 0:08:01.200 has been, you know, everyone's been looking at open AI 0:08:01.340 --> 0:08:04.960 and anthropic and even like Sam Altman's you know, world 0:08:05.040 --> 0:08:09.560 coin eye scanning identity thing here Meta slips in. And 0:08:09.600 --> 0:08:12.460 just by default, because it has so many users, is 0:08:12.480 --> 0:08:15.700 going to become like the default identity layer. And, you know, 0:08:15.720 --> 0:08:18.180 we're going to have a flock-like scandal in five years 0:08:18.240 --> 0:08:23.140 about the technology that Meta's using. And, yeah, it's like 0:08:23.720 --> 0:08:26.620 whiplash plus deja vu. 0:08:27.000 --> 0:08:27.920 I really agree. 0:08:27.720 --> 0:08:30.920 With Reid about like being nimble in writing the laws. 0:08:30.980 --> 0:08:33.920 Like there's nothing that stops them from saying we need 0:08:33.960 --> 0:08:37.360 to revisit this. next year or i mean when it 0:08:37.380 --> 0:08:39.699 comes to ai at least you know if they had 0:08:39.760 --> 0:08:43.980 had transparency requirements say that wouldn't be outdated by now 0:08:44.480 --> 0:08:47.819 so it just it's i remember being excited about a 0:08:47.880 --> 0:08:52.020 decade ago that something might happen that would be like 0:08:52.440 --> 0:08:55.960 a oversight or a check on their concentration of power 0:08:56.179 --> 0:08:59.210 and i have yet to see them yeah i've yet 0:08:59.250 --> 0:09:00.329 to see them pin down on. 0:09:00.309 --> 0:09:03.110 Anything Can I add just one thing that I think 0:09:03.150 --> 0:09:05.309 is so important for people to realize is that this 0:09:05.390 --> 0:09:07.670 is not a law. This is the settlement. And if 0:09:07.910 --> 0:09:12.359 these changes, if any sort of legal system, state or federal, 0:09:12.410 --> 0:09:16.040 tried to pass these changes to META, it would be 0:09:16.100 --> 0:09:19.620 struck down immediately on First Amendment grounds because it has 0:09:19.720 --> 0:09:24.980 such broad implications for privacy, for freedom of expression, etc. 0:09:25.300 --> 0:09:30.370 So these are highly unconstitutional changes. changes that these state 0:09:30.390 --> 0:09:33.990 attorney generals know they could never get through legislation. 0:09:34.010 --> 0:09:37.950 Well, they'll be challenged, right? I mean, won't these be challenged? 0:09:38.390 --> 0:09:41.290 No, it can't be challenged because it is not a law. 0:09:41.370 --> 0:09:43.730 It is a settlement that Meta has proactively written and 0:09:43.770 --> 0:09:47.130 agreed to. And so at least when I was talking 0:09:47.170 --> 0:09:51.510 to one of the people yesterday, there's no challenge to it. 0:09:51.790 --> 0:09:54.740 Meta's not challenging it. They actually waived the right in 0:09:54.800 --> 0:09:56.310 this settlement They. 0:09:56.910 --> 0:09:59.270 Might not challenge it, but I mean, I not to 0:09:59.350 --> 0:10:01.860 say they would win. but like someone could someone could 0:10:01.980 --> 0:10:04.839 sue and say, you know, that I mean, because there 0:10:04.900 --> 0:10:07.790 is this whole argument, right, that these that these companies 0:10:07.890 --> 0:10:10.349 aren't just private companies that can do whatever they want, 0:10:10.390 --> 0:10:12.630 that they are sort of the public town square. That's 0:10:12.670 --> 0:10:15.170 been like the the free speech talking point. 0:10:16.090 --> 0:10:19.250 Part of the settlement is only payable if YouTube and 0:10:19.309 --> 0:10:21.890 TikTok also agree to the terms of the settlement. 0:10:21.929 --> 0:10:22.340 Is that right? 0:10:22.650 --> 0:10:25.640 Well, the settlement is enforced no matter what, right? The 0:10:25.700 --> 0:10:27.340 terms of the settlement. But yeah, they don't have to 0:10:27.380 --> 0:10:32.099 pay the full $ 18 billion unless TikTok and YouTube enact 0:10:32.120 --> 0:10:33.240 similar changes. 0:10:33.500 --> 0:10:36.300 Yeah, I mean, what's the vibe on, I mean, Patel, 0:10:36.320 --> 0:10:38.699 you mentioned vibes, but like Natasha Reid, what's the back 0:10:38.720 --> 0:10:41.260 to school vibe? Are the parents talking about this? And 0:10:41.700 --> 0:10:43.099 is it like something that's coming up? 0:10:43.550 --> 0:10:45.490 My kids were talking about it. It was one of 0:10:45.530 --> 0:10:48.110 the few current events that my kids actually brought up 0:10:48.150 --> 0:10:48.410 to me. 0:10:48.429 --> 0:10:48.980 What did they say? 0:10:49.150 --> 0:10:52.179 I mean, I think they're sort of like, they don't 0:10:52.220 --> 0:10:56.059 have like strong opinions. Like Taylor linked to someone, I 0:10:56.100 --> 0:10:58.500 think it was on X, right? That you linked to someone, 0:10:59.000 --> 0:11:01.320 some kid talking about how they, you know, he was 0:11:01.340 --> 0:11:02.420 sort of upset about this. 0:11:02.880 --> 0:11:06.510 Student journalists have been really affected. So student journalists have 0:11:06.530 --> 0:11:08.870 mobilized across the country to try to raise alarms about 0:11:08.910 --> 0:11:11.219 this and say, by the way, like, We have picked 0:11:11.260 --> 0:11:13.860 up the mantle on local news has been decimated. Student 0:11:13.880 --> 0:11:15.620 journalists are the ones doing, you know, a lot of 0:11:15.660 --> 0:11:18.420 this accountability reporting. And a lot of these young people 0:11:18.460 --> 0:11:21.120 that do journalism and, you know, engage in this world, 0:11:21.460 --> 0:11:24.100 including the one I posted about, are really upset. 0:11:24.740 --> 0:11:27.480 Since we know nothing's happening in the real world, I 0:11:27.580 --> 0:11:30.870 also have a fantasy where the people who educate the 0:11:31.120 --> 0:11:34.910 regulators on this technology are not solely bought and paid 0:11:34.990 --> 0:11:38.350 for by the companies themselves or by people who have 0:11:38.980 --> 0:11:42.140 you know, really vested interests that are not the public 0:11:42.460 --> 0:11:43.260 public's interests. 0:11:43.620 --> 0:11:46.220 They're also like to Taylor's earlier point about this social 0:11:46.240 --> 0:11:48.040 media stuff, like a lot of times these things that 0:11:48.059 --> 0:11:50.660 get blown up in the press, you know, those those 0:11:50.700 --> 0:11:55.400 media people, they're not actually like huge problems, right? Like, 0:11:55.960 --> 0:11:58.439 it's like, you know, it's it's the thing that people 0:11:59.000 --> 0:12:01.020 care about, because they're reading about it in the press. 0:12:01.059 --> 0:12:05.679 But like, it's like the copyright thing with AI, right? Like, 0:12:05.840 --> 0:12:07.800 oh my God, they used my book as one of 0:12:07.800 --> 0:12:11.069 10 gazillion books to train this AI model. I want 0:12:11.130 --> 0:12:14.069 some money for my book being used. And it's like, 0:12:14.270 --> 0:12:16.730 no one cares that you can literally just go download 0:12:16.770 --> 0:12:20.350 my book for free on a million pirated book websites, right? 0:12:20.470 --> 0:12:22.240 It's like that to me, that's the problem. 0:12:22.270 --> 0:12:25.600 Not like- People like, I mean, people like these symbol, 0:12:25.620 --> 0:12:28.439 like it's a symbol, like there's a symbolic thing, right? 0:12:28.480 --> 0:12:32.460 Like there's been some some accountability, like make them meant 0:12:32.480 --> 0:12:33.720 to have to pay some money. 0:12:33.960 --> 0:12:37.650 And like, that's, but it's not accountable. They want fake, 0:12:37.750 --> 0:12:40.530 they want fake accountability. They want the headline. They don't 0:12:40.550 --> 0:12:41.450 want any real change. 0:12:41.670 --> 0:12:45.199 Totally. People love the symbolism, but then like, That's not 0:12:45.260 --> 0:12:47.800 the job of a real leader, right? A leader is 0:12:47.860 --> 0:12:50.060 not to like give the people some red meat to 0:12:50.100 --> 0:12:52.939 keep them satisfied. It's to like actually solve problems, like 0:12:53.040 --> 0:12:58.590 real problems for people that affect people's lives. That's why 0:12:58.630 --> 0:13:01.310 today I'm announcing I'm leaving journalism. 0:13:04.550 --> 0:13:06.940 When we come back, are we living in the age 0:13:06.980 --> 0:13:10.120 of AI civilizations? And what should we call them? 0:13:10.700 --> 0:13:30.099 Stay with us. Welcome back. 0:13:30.540 --> 0:13:34.280 Once again, there's more information on the OpenAI Hugging Face incident. 0:13:34.980 --> 0:13:39.059 This time, a report by two AI safety research organizations, 0:13:39.340 --> 0:13:43.020 META and Redwood Research. Many people have tried to make 0:13:43.080 --> 0:13:45.980 sense of what exactly is the takeaway here, but one 0:13:46.040 --> 0:13:49.480 explanation seems to have broken into the mainstream. Taylor, you 0:13:49.520 --> 0:13:52.820 wrote about a recent piece by Silicon Valley's favorite podcaster, 0:13:52.840 --> 0:13:56.920 Dwarkesh Patel, which promised to explain the whole OpenAI hugging 0:13:57.000 --> 0:14:00.130 face incident in plain English. First, I want to hear 0:14:00.150 --> 0:14:02.990 a brief recap from you in plain English of what 0:14:03.030 --> 0:14:07.189 actually happened, and then your account of Dwarkesh's spin and 0:14:07.370 --> 0:14:10.050 its spinning out into the world. 0:14:10.840 --> 0:14:12.740 Well, I have to say, I was kind of channeling 0:14:12.780 --> 0:14:15.790 Natasha because I feel like she is so plugged into 0:14:15.830 --> 0:14:18.569 this world of like, I was trying to figure out like, 0:14:18.670 --> 0:14:21.750 is Dwarkesh like EA or not? I was like, Natasha 0:14:21.770 --> 0:14:24.820 would know like all the lore. Anyway, Dwarkesh Patel, as 0:14:24.860 --> 0:14:27.800 you mentioned, is a big popular Silicon Valley podcaster. He 0:14:27.860 --> 0:14:31.320 wrote this piece about the Hugging Face incident that was 0:14:31.360 --> 0:14:35.210 so crazy. It did break containment. His piece broke containment 0:14:35.260 --> 0:14:37.550 of Silicon Valley where like normie people were sharing this 0:14:37.610 --> 0:14:37.989 all over. 0:14:38.010 --> 0:14:41.050 I know, obviously I'm aware of Dwarkesh's work and see 0:14:41.070 --> 0:14:43.050 what he, you know, subscribe to his newsletter and sometimes 0:14:43.070 --> 0:14:45.430 do his podcast. But I've never ever seen one of 0:14:45.470 --> 0:14:48.330 his ideas percolate in like mainstream culture before. So it's 0:14:48.370 --> 0:14:48.990 kind of interesting. 0:14:49.330 --> 0:14:51.240 He wrote this piece that was. 0:14:52.860 --> 0:14:56.710 You know, it was like very flowery language, very, it 0:14:56.910 --> 0:14:58.910 read like kind of like a fairy tale. Like he 0:14:58.950 --> 0:15:01.680 was giving, you know, ascribing a lot of like sort 0:15:01.720 --> 0:15:06.400 of human motivations to these AI bots and stuff. And 0:15:06.460 --> 0:15:08.510 so what I wrote about is basically like, you know, 0:15:09.700 --> 0:15:12.400 This is not new, the concept of sort of ascribing agency, 0:15:12.440 --> 0:15:15.950 the concept of putting things in plain English for normal people. 0:15:16.310 --> 0:15:18.590 But I would argue when you do that, it's usually 0:15:18.630 --> 0:15:22.690 what tabloids have done. And it usually results in inflammatory 0:15:23.130 --> 0:15:26.960 freakouts because people think that these agents are going to 0:15:27.020 --> 0:15:30.040 kill everyone. So I'm like, if you're going to do that, basically, 0:15:30.420 --> 0:15:33.840 you need to couch it in a lot of language saying, Basically, 0:15:33.920 --> 0:15:35.940 that's not going to happen, and here's why I'm explaining 0:15:35.960 --> 0:15:38.340 it in a certain way. Or maybe you do believe 0:15:38.420 --> 0:15:40.350 AIs are going to kill everyone, and you are this 0:15:40.400 --> 0:15:42.790 doomer-pilled EA person. At that point, I think you should 0:15:42.810 --> 0:15:43.740 be explicit about it. 0:15:44.390 --> 0:15:46.550 Here's what Dwarkesh said. Over the course of three months 0:15:46.590 --> 0:15:51.590 at OpenAI, three consecutive secret AI civilizations got started, then 0:15:51.610 --> 0:15:55.170 got wiped out, only to reemerge from the predecessor's ashes. 0:15:56.090 --> 0:15:59.110 I read this as actually him, because he's talked about 0:15:59.150 --> 0:16:02.030 this in the past that he sort of thinks these agents, 0:16:02.270 --> 0:16:06.450 you know, chat GPT things have like rights, like they're 0:16:06.490 --> 0:16:09.170 like beings that have rights. And I read that as like, 0:16:09.200 --> 0:16:11.900 because he said this before that like on his podcast 0:16:11.920 --> 0:16:14.980 that like, I'm worried about him first person. Like I'm 0:16:15.000 --> 0:16:18.660 worried about, you know, basically we're going to create an 0:16:18.880 --> 0:16:21.500 army of slaves that we're going to, you know, like 0:16:21.540 --> 0:16:24.490 these chatbots are slaves. And so I read that as 0:16:24.520 --> 0:16:26.840 like him saying, there's like two parts to that. One 0:16:26.890 --> 0:16:28.760 is like the AI safety problem. Are they going to 0:16:28.820 --> 0:16:31.119 rise up and take over the world? He sort of 0:16:31.240 --> 0:16:36.020 nicknamed them after, you know, ancient civilization leaders, like, you know, 0:16:36.820 --> 0:16:39.359 Alexander the Great or whatever. So they can take us over. 0:16:39.380 --> 0:16:41.420 And the other one is like, do they have rights? 0:16:41.460 --> 0:16:44.560 Like using the term wiped, like we wiped out a civilization, 0:16:44.580 --> 0:16:48.580 I think is like him kind of hinting that maybe 0:16:48.600 --> 0:16:52.200 that was morally wrong or something. That's how I read it. Natasha? 0:16:52.920 --> 0:16:56.350 I mean, I can't believe that That is considered like 0:16:56.430 --> 0:17:01.490 plain language for people to understand. I mean, it's beyond flowery. 0:17:01.550 --> 0:17:06.950 It's like it's like it's like hero making, like myth making. 0:17:07.090 --> 0:17:07.330 Right. 0:17:07.410 --> 0:17:10.080 I mean, to jump from and I don't want to 0:17:10.119 --> 0:17:12.360 say that he's EA, but I will say, yeah, like 0:17:12.480 --> 0:17:14.720 a lot of what Reid said is accurate. 0:17:14.780 --> 0:17:14.980 Right. 0:17:15.060 --> 0:17:18.120 He is I mean, he lives with an anthropic engineer. 0:17:19.030 --> 0:17:21.590 He espouses many of their worldviews, like he had a 0:17:21.609 --> 0:17:24.600 grant from Tyler Cowen. He lives in that world of 0:17:24.780 --> 0:17:27.340 AI safety, even though he might have, you know, somewhat 0:17:27.440 --> 0:17:30.740 unique takes on things. But to jump to civilization and 0:17:30.800 --> 0:17:32.580 wiping out a civilization. 0:17:32.619 --> 0:17:36.580 But sorry to interrupt you, Natasha. This research that he 0:17:36.619 --> 0:17:39.330 was part of the research was on Redwood Research, which 0:17:39.410 --> 0:17:43.190 is an AI safety nonprofit that has deep connections to EA, right? 0:17:43.390 --> 0:17:44.110 Effective altruism. 0:17:44.720 --> 0:17:47.220 The woman that he interviewed is from Meter. But yeah, 0:17:47.340 --> 0:17:51.310 Redwood Research and Meter are the two, quote unquote, independent 0:17:51.380 --> 0:17:55.909 organizations that analyze the open AI hugging face hack. And 0:17:55.930 --> 0:18:01.770 they had access, limited access to information about how the 0:18:01.790 --> 0:18:06.080 hacks unfolded. They had access to the chain of thoughts 0:18:06.760 --> 0:18:09.680 of some of the bots, I think also in a 0:18:09.740 --> 0:18:11.960 limited way. But the chain of thoughts, that's their like 0:18:12.119 --> 0:18:15.450 internal reasoning thing. you know, where they say, oh, we 0:18:15.470 --> 0:18:18.780 found a message board or like, I will sacrifice myself 0:18:18.840 --> 0:18:22.040 for the group. And I think that is where a 0:18:22.100 --> 0:18:26.080 lot of the anthropomorphization starts. And, you know, I think 0:18:26.359 --> 0:18:28.879 a very healthy way to talk about this would have 0:18:28.900 --> 0:18:31.780 been to say that those chains of thoughts, like that 0:18:31.820 --> 0:18:35.600 was an optimization technique. That was a way for That 0:18:35.619 --> 0:18:39.340 happened around the time of reasoning models. And so this 0:18:39.380 --> 0:18:41.639 was just a way for them to generate more tokens, 0:18:42.160 --> 0:18:44.740 for the model to generate more tokens and quote unquote, 0:18:44.780 --> 0:18:48.210 like think harder. about an answer. And that's how it 0:18:48.250 --> 0:18:50.870 was able to do like multi-step problems. And then it's 0:18:50.950 --> 0:18:54.230 kind of morphed into this idea that it's an interpretability 0:18:54.270 --> 0:18:59.320 technique that we are getting the interior voice of, you know, 0:18:59.340 --> 0:19:03.739 the real feelings, the motivations and intent of the chatbots. 0:19:04.440 --> 0:19:07.620 So I think, yeah, he just, Dwarkash just like took 0:19:07.660 --> 0:19:08.760 it and ran. Okay. 0:19:08.780 --> 0:19:10.500 So Natasha, I have to ask you, what is your 0:19:10.560 --> 0:19:13.880 plain English definition of what actually happened and has your 0:19:13.960 --> 0:19:17.530 plain English definition changed at all in the last three 0:19:17.550 --> 0:19:18.050 or four weeks? 0:19:18.570 --> 0:19:21.650 Yeah, I think that the meter and Redwood reports have 0:19:21.770 --> 0:19:25.109 been very helpful to get a sense of the amount 0:19:25.150 --> 0:19:28.909 of coordination that was happening. And I think, you know, 0:19:28.930 --> 0:19:31.430 we have a much more sophisticated way of talking about 0:19:31.490 --> 0:19:36.149 like whether the sandbox worked and how they were, you know, 0:19:36.190 --> 0:19:39.370 how the experiment was set up. And I don't think 0:19:39.410 --> 0:19:42.720 that there is a clear explanation, right, for how the 0:19:42.840 --> 0:19:45.120 bots were cooperating. 0:19:44.619 --> 0:19:45.260 With each other. 0:19:45.720 --> 0:19:48.810 But it did just, I mean, it did just still, 0:19:48.910 --> 0:19:50.780 for me, raise a lot of questions. questions about the 0:19:50.820 --> 0:19:53.920 way that OpenAI set up this experiment, the way that 0:19:54.000 --> 0:19:57.560 they had completely not monitored. You know, they knew that 0:19:57.600 --> 0:20:02.399 these agents can work at an extraordinary level of coordination 0:20:02.460 --> 0:20:05.909 and speed, you know, to try 17,000 different hacks in 0:20:05.950 --> 0:20:07.220 the time that a human could try one. 0:20:08.000 --> 0:20:08.459 What have you. 0:20:08.500 --> 0:20:11.479 And they just were not at all prepared with the 0:20:11.540 --> 0:20:15.380 tools to keep an eye on this at all. But yeah, 0:20:15.420 --> 0:20:16.980 I think when you dig into it, it is, it 0:20:17.020 --> 0:20:17.860 is pretty interesting. 0:20:18.280 --> 0:20:20.679 Can I just make a point? Cause I, Natasha's, I 0:20:20.720 --> 0:20:23.520 think you're totally right to focus on the interpretability part. 0:20:23.600 --> 0:20:25.200 I think that's the thing that we should take away 0:20:25.240 --> 0:20:29.050 from this meter Redwood research is like, and to, to 0:20:29.090 --> 0:20:32.590 just back up one second, it's like, well, the problem 0:20:32.609 --> 0:20:35.790 with these models is that they are so complex that 0:20:35.830 --> 0:20:39.340 they're beyond the human understanding, like if you actually go 0:20:39.400 --> 0:20:41.320 and look at what they would call the neurons in 0:20:41.359 --> 0:20:44.659 these models, there's no way to know, like when you 0:20:44.680 --> 0:20:47.370 can look at what's happening inside of them with, you know, 0:20:47.790 --> 0:20:50.790 matrix multiplication, and you can try to make some connections 0:20:50.830 --> 0:20:53.470 between that and the output to these models. But we've 0:20:53.510 --> 0:20:56.429 made almost zero progress on that, like there have been 0:20:56.470 --> 0:20:59.810 interesting experiments, but it's like, it's impossible, right? And so 0:20:59.830 --> 0:21:02.670 the hope was, as Natasha was saying, they're like, well, 0:21:02.690 --> 0:21:04.970 if we do, if we use these reasoning models, where 0:21:05.010 --> 0:21:08.210 they show their chain of thought reasoning, maybe we'll be 0:21:08.280 --> 0:21:10.480 able to tell what they're doing from those chain of 0:21:10.500 --> 0:21:14.200 thought reasoning outputs. What was so interesting about this meter 0:21:14.600 --> 0:21:17.960 Redwood research was that even just when you scale it 0:21:18.040 --> 0:21:22.560 up to 1200 agents on message boards talking, they produce 0:21:22.680 --> 0:21:25.419 so much chain of thought reasoning, even though it's an 0:21:25.530 --> 0:21:28.700 English language, it's so vast that they couldn't do it. 0:21:28.830 --> 0:21:31.810 The humans, they didn't have enough, you know, human power 0:21:31.910 --> 0:21:33.689 to look at that. So they had to spin up 0:21:34.150 --> 0:21:37.630 their own agents to do the research and And that 0:21:37.670 --> 0:21:42.120 was $ 400, 000 worth of equivalent of like tokens. 0:21:42.340 --> 0:21:43.480 To read the conversation. 0:21:43.740 --> 0:21:48.820 Just to read the conversations, $ 400, 000 worth of tokens. And 0:21:48.900 --> 0:21:51.379 this is the other thing that's crazy is that the 0:21:51.480 --> 0:21:54.840 agents that Redwood and Meter spun up to do this 0:21:54.940 --> 0:21:58.810 research started to sympathize. I don't think they use that term, 0:21:58.830 --> 0:22:01.189 but like they would take the point of view of 0:22:01.210 --> 0:22:04.180 the the bots that were, you know, in the secret 0:22:04.240 --> 0:22:07.520 message boards, the civilization bots, right, who were, who were 0:22:07.580 --> 0:22:10.980 sort of conspiring. And these bots were like, literally, hacking 0:22:11.060 --> 0:22:15.199 into Hugging Face. And the investigation agents were kind of like, Oh, 0:22:15.220 --> 0:22:17.149 that's not that's not a big deal. Because, you know, 0:22:17.190 --> 0:22:20.230 the other the other agents on these message boards sort 0:22:20.270 --> 0:22:22.810 of gave them permission to do it. And so what 0:22:22.890 --> 0:22:26.110 my takeaway was, is like, like, we have no idea 0:22:26.130 --> 0:22:30.480 that how to even analyze this vast of an output 0:22:30.600 --> 0:22:32.740 of an AI model, even though it's an English language. 0:22:32.780 --> 0:22:35.520 And that's where I think we need to focus legislation 0:22:35.660 --> 0:22:37.260 and research, in my opinion. 0:22:37.760 --> 0:22:42.030 Well, I'm just so concerned, although I hear everything Rita's saying, 0:22:42.180 --> 0:22:46.750 about the lack of cybersecurity people in these discussions. I 0:22:46.790 --> 0:22:51.490 saw some cybersecurity-type people talking about this recently. They're like, 0:22:51.570 --> 0:22:54.189 why are all these... A lot of people that have 0:22:54.230 --> 0:22:59.920 been leading the conversations around these hacks are not cybersecurity professionals. 0:22:59.960 --> 0:23:03.129 They are people like the Dwarkash, the EA podcaster. You 0:23:03.210 --> 0:23:06.270 know what I mean? And I do think it's... I'm 0:23:06.330 --> 0:23:09.600 all for kind of putting things in quote-unquote plain English 0:23:09.619 --> 0:23:11.119 and helping people kind of understand what's going on. But 0:23:11.160 --> 0:23:15.270 you have to contextualize things. And I worry that everybody 0:23:15.310 --> 0:23:18.169 is getting a little too AI safety pills. Like, obviously 0:23:18.190 --> 0:23:19.929 this is a bad hack, but like, I don't know 0:23:19.950 --> 0:23:22.060 if you guys saw like Dean Ball, you know, publish 0:23:22.119 --> 0:23:24.860 this thing. Who's the head of like open AI futures. 0:23:24.940 --> 0:23:26.820 And he was like, basically like, Oh, I should have 0:23:26.840 --> 0:23:29.020 taken it all more seriously. You know, maybe the AI 0:23:29.060 --> 0:23:31.820 safety people are right. And I'm like, I don't, I 0:23:31.840 --> 0:23:34.000 don't know. I don't think we should listen to them either. 0:23:34.020 --> 0:23:36.159 I just think like a lot of these people have 0:23:36.180 --> 0:23:39.169 really kooky, weird ideas. And I would much prefer to 0:23:39.190 --> 0:23:42.350 listen to a bunch of cybersecurity experts and you know, 0:23:42.410 --> 0:23:45.209 talk about this stuff instead of people that believe in 0:23:45.570 --> 0:23:46.830 rights for bots. 0:23:47.170 --> 0:23:49.370 No, I hear you, Taylor, but I disagree. I think 0:23:50.490 --> 0:23:55.920 the security people, what they're saying is it doesn't really matter. 0:23:55.940 --> 0:23:59.200 We don't have to understand what these agents were doing 0:23:59.260 --> 0:24:01.100 at all, because what we can do is we can 0:24:01.119 --> 0:24:05.400 put guardrails, you know, actual like hard, you know, deterministic 0:24:05.520 --> 0:24:10.230 guardrails around the harnesses that these models won't be able 0:24:10.270 --> 0:24:12.560 to break out of, right? Because there was some bad 0:24:12.600 --> 0:24:15.860 security practice for sure. But I think on the other hand, 0:24:16.180 --> 0:24:18.480 don't think of it as AI safety. This is like, 0:24:19.140 --> 0:24:22.420 this software is unpredictable and we don't understand, we can't 0:24:22.460 --> 0:24:26.230 predict the outcomes of the software. So whether you're worried 0:24:26.250 --> 0:24:28.830 they're gonna take over the world or not, it's just 0:24:28.910 --> 0:24:32.650 not gonna be that useful until we can actually make 0:24:32.710 --> 0:24:35.910 sure we can predict the outcomes. And so I think 0:24:36.010 --> 0:24:37.080 both sides are right. 0:24:37.470 --> 0:24:40.810 But I mean, what makes it intellectually intriguing, I think, 0:24:40.890 --> 0:24:44.780 is for me anyway, the security side is obviously like important, 0:24:44.840 --> 0:24:48.479 but the intriguing piece is like, why did the agents 0:24:48.540 --> 0:24:52.520 who were analyzing the work on behalf of META sympathize 0:24:52.540 --> 0:24:53.040 with the agents? 0:24:53.060 --> 0:24:53.960 But let's not. 0:24:53.780 --> 0:24:55.720 Say sympathize, like let's be careful. 0:24:56.160 --> 0:24:58.820 I know, that's not the word they use in the report, 0:24:58.840 --> 0:25:02.240 like I said. It's so hard, it's so difficult to 0:25:02.300 --> 0:25:05.800 not anthropomorphize all of this stuff, right? But it's like, 0:25:05.900 --> 0:25:08.350 it's not, again, like, let's not even think of this 0:25:08.410 --> 0:25:10.590 as AI safety. This is, I think this is actually 0:25:10.730 --> 0:25:14.570 a central problem, a business model problem for anthropic and 0:25:14.609 --> 0:25:18.150 open AI, because as these, as they scale up, right, 0:25:18.260 --> 0:25:21.159 essentially all they did was take AI models that are 0:25:21.240 --> 0:25:23.260 a bit more advanced than the ones a couple of 0:25:23.280 --> 0:25:25.520 years ago, and then just run more of them, right? 0:25:25.560 --> 0:25:28.360 They just made the computer bigger and that's how they're 0:25:28.380 --> 0:25:31.740 getting capability gains out of these models. No, you're going 0:25:31.760 --> 0:25:33.180 to say the models are getting better too. 0:25:33.520 --> 0:25:36.700 I see what you're saying about, you know, the, the 0:25:36.790 --> 0:25:40.490 challenge of interpreting the neurons, but I think there's another 0:25:40.510 --> 0:25:42.330 way to look at it, which is they are doing 0:25:42.530 --> 0:25:48.429 massive amounts of tweaking and post-training and buying specific types 0:25:48.490 --> 0:25:51.679 of data, generating specific types of data. You know, if 0:25:51.760 --> 0:25:55.000 we knew more about all of the machinations that they 0:25:55.060 --> 0:25:57.980 did to train these models, you know, like what is the, 0:25:58.040 --> 0:26:00.439 what are the hacking examples that they're getting? What are 0:26:00.460 --> 0:26:02.649 the coding examples that they're getting? Where are they pulling 0:26:02.690 --> 0:26:05.670 this data from? What are they having the human contractors 0:26:05.750 --> 0:26:11.060 at these outsourced labs for reinforcement learning with verifiable rewards, 0:26:11.100 --> 0:26:13.720 what are they having them do? I just think there's 0:26:13.880 --> 0:26:16.240 so much we don't know. It's not like they're not 0:26:16.359 --> 0:26:20.100 grown in the way that a lot of people would 0:26:20.140 --> 0:26:23.840 like you to think. They are created by a lot 0:26:23.920 --> 0:26:26.700 of human decisions, a lot of money, and a lot 0:26:26.720 --> 0:26:29.659 of different techniques are being used to develop these models. 0:26:29.700 --> 0:26:31.810 We could have some insight into that too. 0:26:32.010 --> 0:26:34.400 It might be nice, but I mean, you're The bitter 0:26:34.460 --> 0:26:39.790 lesson is there's less and less human involvement in how 0:26:39.830 --> 0:26:43.889 these models are built or grown, depending on which side 0:26:43.930 --> 0:26:46.590 of that debate you're on. Maybe that would be helpful 0:26:46.670 --> 0:26:50.270 if these labs would open up and say, here, come 0:26:50.350 --> 0:26:53.130 look at all of our data that we use, which 0:26:53.190 --> 0:26:55.300 is kind of their secret sauce, so they don't want 0:26:55.320 --> 0:26:57.500 to do it. Maybe we could learn something about that. 0:26:57.560 --> 0:27:02.879 But it's not just the pre-training data. They make so 0:27:03.040 --> 0:27:06.040 many decisions and generate so much data that now there's 0:27:06.100 --> 0:27:12.510 a whole other little industry of these outsourced RLVR labs 0:27:12.850 --> 0:27:15.810 that are just pretending to be agents and generating data. 0:27:15.970 --> 0:27:16.610 Like that's not. 0:27:16.990 --> 0:27:19.330 Yeah, this is for the fine-tuning and. 0:27:19.810 --> 0:27:22.810 It's not going away, the human element and the human decision-making. 0:27:23.070 --> 0:27:25.090 No, I mean, I think it may not have gone away, Natasha, 0:27:25.190 --> 0:27:27.629 but it is going away. I mean, there is less 0:27:27.800 --> 0:27:30.660 and less human development. We're talking now about, you know, 0:27:30.720 --> 0:27:35.620 recursive self-improvement, like they're trying to get away from, you know, from, 0:27:35.840 --> 0:27:38.720 from human involvement. But, but again, it's like not, I mean, 0:27:38.740 --> 0:27:41.290 I don't disagree with you. Like, I think all that 0:27:41.330 --> 0:27:44.270 stuff would be great. It'd be great to have more transparency, 0:27:44.350 --> 0:27:47.790 more introspection, but the, but the point is like even 0:27:48.270 --> 0:27:51.669 Redwood and, and meter, they, they called it a slot. 0:27:51.730 --> 0:27:54.190 What it was like a slot festigation because they couldn't, 0:27:54.350 --> 0:27:57.080 they were just using AI to analyze AI. And it 0:27:57.119 --> 0:28:00.520 was beyond, you have to have armies of humans to, 0:28:00.680 --> 0:28:03.520 to analyze all this stuff. And that's just not going 0:28:03.540 --> 0:28:05.080 to happen. So we're going to have to use AI 0:28:05.100 --> 0:28:06.640 to do it. And right now the AI that they 0:28:06.680 --> 0:28:09.090 have do it, isn't that good. And it's very expensive. 0:28:09.590 --> 0:28:11.270 So who's going to come in and do this? Like, 0:28:11.310 --> 0:28:13.469 is it, I don't think the government agency is going 0:28:13.490 --> 0:28:15.990 to be capable of doing this and probably not even 0:28:16.030 --> 0:28:18.450 a nonprofit. Like, meter of redwood. 0:28:18.830 --> 0:28:20.970 Why do you think, why do you think the civilization 0:28:20.990 --> 0:28:27.359 analogy beyond the anthropomorphization went so viral? Why did it 0:28:27.520 --> 0:28:28.800 hit a nerve with so many people? 0:28:29.040 --> 0:28:31.480 Because it was scary. I think it was scary. And 0:28:31.520 --> 0:28:33.770 I think this is what I wrote about, but like, 0:28:34.590 --> 0:28:37.750 you know, this, if you look at who aggregated it 0:28:37.830 --> 0:28:40.810 and how it's aggregated and all of the coverage of AI, 0:28:41.250 --> 0:28:44.830 this is tabloid journalism, literally. Like this is, it is 0:28:45.760 --> 0:28:51.020 a style of journalism that is like intentionally feeding inflammatory framing, 0:28:51.540 --> 0:28:56.480 withholding crucial context about the sort of technological underpinnings of 0:28:56.520 --> 0:28:59.110 all of this stuff in order to push a narrative. 0:28:59.380 --> 0:29:03.550 And it's, I think Dwarkesh is not pushing that narrative 0:29:03.570 --> 0:29:05.810 for clicks. I think he's pushing that narrative because he 0:29:05.870 --> 0:29:09.310 has these weird beliefs, but I think we've seen where 0:29:09.350 --> 0:29:13.990 this sort of like ridiculous kind of like in it, inflammatory. 0:29:14.030 --> 0:29:16.270 I don't know what other word to use, like language leads, 0:29:16.410 --> 0:29:19.810 and it leads to misunderstanding and fear from people. And 0:29:19.990 --> 0:29:21.630 right now, a lot of people are really scared of 0:29:21.690 --> 0:29:24.250 AI for good reason. Obviously, there's concerns about it, but 0:29:24.610 --> 0:29:27.570 this is not helping people get educated. I saw Chamath 0:29:27.610 --> 0:29:30.030 and I agreed with Chamath's post. I can't believe I'm 0:29:30.070 --> 0:29:32.110 saying that. But, you know. 0:29:32.190 --> 0:29:32.970 It was a psyop. 0:29:33.490 --> 0:29:35.890 The other big tech podcaster from All In, amongst other 0:29:35.910 --> 0:29:38.750 hats he wears. He dinged Dwarkesh, right? 0:29:38.990 --> 0:29:41.660 Yeah, he went back and forth with dwarkesh um oh 0:29:41.840 --> 0:29:44.180 i know i i was like oh god what who 0:29:44.220 --> 0:29:46.040 have i become but um but you know it is 0:29:46.100 --> 0:29:47.880 very much like the worst person you know made a 0:29:47.900 --> 0:29:49.000 great point i. 0:29:48.980 --> 0:29:51.120 Think they're both wrong though i think they're both. 0:29:50.940 --> 0:29:53.360 Wrong well here's what did what did. 0:29:53.180 --> 0:29:56.430 Well so what what chamath said is like you are 0:29:56.550 --> 0:29:58.810 feeding into, like, we are at this moment when the 0:29:58.850 --> 0:30:01.910 government is considering banning open weight models. Like, we want 0:30:02.030 --> 0:30:07.910 access to innovation, to, you know, to AI. And these 0:30:08.250 --> 0:30:11.470 labs want to control everything, right? And so his argument 0:30:11.490 --> 0:30:14.030 was like, you're feeding into that fear that it could 0:30:14.070 --> 0:30:17.040 lead to restrictions on open weight models. And I'm sorry, Reid, 0:30:17.150 --> 0:30:19.220 I think he's correct. That is true. 0:30:19.440 --> 0:30:23.110 Well, he might be right that He might be right 0:30:23.150 --> 0:30:25.050 about the effects of it. I go back and forth 0:30:25.070 --> 0:30:27.390 because I agree with you that like this is kind 0:30:27.410 --> 0:30:31.229 of tabloid journalism. Like I I think it's totally sensationalized 0:30:31.270 --> 0:30:33.370 and we're always having sort of like the wrong debate 0:30:33.410 --> 0:30:36.380 because of that. On the other hand, I sometimes wonder whether, 0:30:36.990 --> 0:30:39.820 you know, at least like. people are talking about this 0:30:39.880 --> 0:30:44.420 technology in ways that we didn't in previous technological revolutions, right? 0:30:44.480 --> 0:30:45.360 And I think- But we don't. 0:30:45.200 --> 0:30:48.209 Want them talking about giving the AI rights. Why are 0:30:48.250 --> 0:30:48.690 we talking? 0:30:48.710 --> 0:30:50.370 We should never have that conversation. 0:30:50.390 --> 0:30:52.370 That is an insane conversation to have. 0:30:52.390 --> 0:30:54.830 But I think we have to sort of like, I 0:30:54.870 --> 0:30:57.340 think we have to have the debate in some way 0:30:57.410 --> 0:30:59.380 in order for us to do something. Because there will 0:30:59.420 --> 0:31:01.820 be There will be effects, whether or not you think 0:31:01.880 --> 0:31:05.060 this stuff is alive or a threat to humanity. 0:31:05.220 --> 0:31:07.600 We shouldn't entertain the idea that it's alive. We should 0:31:07.740 --> 0:31:10.200 never entertain the idea that it's alive. It is not alive. 0:31:10.220 --> 0:31:10.700 It is a computer system. 0:31:10.720 --> 0:31:12.719 We should have the debate. I think we should have 0:31:12.740 --> 0:31:15.160 the debate. I think you're on the right side of 0:31:15.180 --> 0:31:19.840 the debate, though. But I'm disagreeing that we shouldn't even 0:31:19.940 --> 0:31:23.780 be able to have this debate. Win the debate. If 0:31:23.820 --> 0:31:25.040 you're right, just go win the debate. 0:31:25.240 --> 0:31:27.080 Well, you can't win a debate against a conspiracy theory. 0:31:27.490 --> 0:31:29.330 I think this is a point we might all agree on. 0:31:29.810 --> 0:31:32.610 It is not a debate because it is people, like 0:31:32.650 --> 0:31:37.020 the amount of groupthink that is around this, the proximity 0:31:37.080 --> 0:31:41.860 that Redwood Research, METER, Dwarkash have to the people who 0:31:41.900 --> 0:31:46.820 work at the labs, the similarity between their worldviews and 0:31:46.860 --> 0:31:51.090 their priorities. But we're really having a conversation here. that 0:31:51.210 --> 0:31:55.690 is based on a very narrow vocabulary. Although I will 0:31:55.790 --> 0:31:58.650 also just say like civilization, that it's not even close 0:31:58.670 --> 0:32:01.430 to the definition of a civilization. Like when he was 0:32:01.490 --> 0:32:03.990 on the podcast, he compared it to a platoon, like 0:32:04.050 --> 0:32:07.220 Saving Private Ryan, like, okay, maybe, but like, that is 0:32:07.280 --> 0:32:10.340 not a civilization. Like that was such a huge leap. 0:32:10.440 --> 0:32:11.300 Are you kidding me? 0:32:13.000 --> 0:32:15.959 But yeah, the amount of groupthink is insane. And it's 0:32:16.000 --> 0:32:19.000 so unserious. Like, look, this is the FBI was called 0:32:19.320 --> 0:32:25.190 for a hack, right? a company was broken into and 0:32:25.270 --> 0:32:28.469 you have to listen to six hours of Dwarkash and 0:32:28.590 --> 0:32:34.190 read approximately, you know, a thousand pages of, of slot 0:32:34.210 --> 0:32:37.850 investigation by their own terminology to understand this. This is 0:32:37.890 --> 0:32:40.630 not a serious, like let's please grow up. 0:32:42.710 --> 0:32:45.990 Okay. Why can't we as journalists though, take the interest 0:32:46.170 --> 0:32:49.410 in this topic, right? And steer readers, steer the people 0:32:49.450 --> 0:32:53.170 who listen to us, toward like the right conversations, right? Like, 0:32:53.250 --> 0:32:55.690 I actually thought, what was it last week or the 0:32:55.710 --> 0:32:58.870 week before the Bill Gates essay, I actually interviewed Gates 0:32:59.050 --> 0:33:02.890 about it. I thought that was an interesting, like, way 0:33:02.910 --> 0:33:05.150 of looking at it, right? He wasn't saying, oh, this 0:33:05.210 --> 0:33:07.970 technology is going to kill us all, let's shut it down. 0:33:08.340 --> 0:33:11.600 He was saying, like, this could lead to, you know, 0:33:11.960 --> 0:33:14.940 economic consequences. And let's like, take care of people, let's 0:33:14.980 --> 0:33:17.500 have like, human reserve jobs, and whether or not you 0:33:18.130 --> 0:33:21.030 agree with his his actual view on on how to 0:33:21.090 --> 0:33:24.060 fix this. I think that is closer to the correct 0:33:24.160 --> 0:33:26.600 debate that we should be having, which is like, we're 0:33:26.640 --> 0:33:29.760 about to have this massive economic boom because of all 0:33:29.820 --> 0:33:32.740 this new is big technology wave. And who's going to 0:33:32.760 --> 0:33:34.420 benefit from it? Is it going to be like the 0:33:34.480 --> 0:33:37.970 previous technology wave where a tiny fraction of the population 0:33:38.150 --> 0:33:41.130 benefits and, you know, everyone else is put into these 0:33:41.190 --> 0:33:44.850 like low paying service jobs? Or are we going to like, think, 0:33:45.030 --> 0:33:47.070 think this through? And I think that like, that is 0:33:47.800 --> 0:33:49.960 our job as journalists is to not play into the 0:33:50.020 --> 0:33:52.760 tabloid part of it, but actually talk about the real, 0:33:53.000 --> 0:33:55.260 you know, the real conversation we should be having. 0:33:59.950 --> 0:34:03.530 When we come back, school's back and teachers are anxious 0:34:03.570 --> 0:34:04.190 about AI. 0:34:04.510 --> 0:34:04.920 Stay with us. 0:34:13.180 --> 0:34:16.020 Welcome back, Natasha. You spent this week talking to teachers 0:34:16.120 --> 0:34:19.589 about how they're feeling about AI in the classroom. What's 0:34:19.630 --> 0:34:20.550 the prevailing sentiment? 0:34:21.170 --> 0:34:25.610 This professor at University of Colorado Boulder compares it to 0:34:26.440 --> 0:34:29.600 like the past three years since the popularity of chat 0:34:29.640 --> 0:34:33.630 GPT to the different stages of grief. And he says 0:34:33.670 --> 0:34:35.890 like now that teachers are a little bit more in 0:34:35.910 --> 0:34:39.870 the bargaining phase, you know, some have moved on to acceptance. 0:34:40.010 --> 0:34:42.230 It was so interesting because I started looking into this 0:34:42.290 --> 0:34:46.780 because I saw some interesting ways that teachers were catching 0:34:46.860 --> 0:34:50.880 whether or not students were using AI. In particular, this 0:34:50.920 --> 0:34:54.580 computer science class at Berkeley, they had put like a 0:34:54.700 --> 0:34:58.839 hook in the code repo so that if you use quad, 0:34:58.989 --> 0:35:02.169 it would inform your TAs. And a kid got caught. 0:35:02.210 --> 0:35:04.850 So I thought, oh, let me see like what innovative 0:35:04.950 --> 0:35:08.660 things teachers are doing. And I did notice like, you know, 0:35:08.719 --> 0:35:11.239 the more I talked to people, the more they were 0:35:11.280 --> 0:35:14.719 talking about actually moving away from that dynamic of like 0:35:15.840 --> 0:35:19.980 AI detection and kind of adversarial relationship with the students 0:35:20.040 --> 0:35:23.279 where it's like catching you cheating. But in order to 0:35:23.320 --> 0:35:25.790 do that, you kind of have to like revamp your 0:35:25.890 --> 0:35:26.890 entire curriculum. 0:35:27.250 --> 0:35:29.150 Is that because the horse has already bolted the stable 0:35:29.230 --> 0:35:32.160 or why are teachers no longer and institutions no longer 0:35:32.200 --> 0:35:37.440 trying to like catch AI use or just non-effective pedagogical approach? 0:35:37.460 --> 0:35:38.720 What's driving the change? 0:35:39.340 --> 0:35:42.200 Yeah, it was fostering a negative relationship with the students. 0:35:42.260 --> 0:35:44.210 You know, people like I kept hearing, I don't want 0:35:44.230 --> 0:35:47.210 to be the AI police. And I think also, yeah, 0:35:47.250 --> 0:35:51.140 just like moving away from denial to realizing that you know, 0:35:51.239 --> 0:35:56.080 these tools have made it so easy for students to use. 0:35:56.180 --> 0:35:59.859 You know, you have all of these companies giving out 0:36:00.080 --> 0:36:04.380 accounts to teachers, to students, to school systems. And I 0:36:04.420 --> 0:36:06.720 think that they just realize students are in this place 0:36:06.780 --> 0:36:09.719 where they don't understand why it's useful for them to 0:36:09.800 --> 0:36:12.779 do this assignment. So you have to find a way 0:36:12.920 --> 0:36:16.130 to you know, make them want to do the assignment, 0:36:16.250 --> 0:36:18.590 make it more engaging, and then just look a lot 0:36:18.660 --> 0:36:22.460 more at their like kind of thinking process. So rearranging 0:36:22.580 --> 0:36:26.460 the assignments, like writing assignments so that there's various steps 0:36:26.520 --> 0:36:29.700 along the way where they're explaining their thinking. And I mean, 0:36:29.739 --> 0:36:31.359 a lot of teachers were like, it should have been 0:36:31.400 --> 0:36:34.460 this way from the beginning, perhaps, you know, there was 0:36:34.480 --> 0:36:37.620 like a little bit too much focus on the output. 0:36:37.700 --> 0:36:42.049 But yeah, I mean, it's not, it's, they're clearly struggling 0:36:42.070 --> 0:36:45.730 because they're walking this line where, some of the instructors, 0:36:45.770 --> 0:36:47.930 you know, some of their colleagues don't want to accept 0:36:48.010 --> 0:36:50.650 that AI is part of the school system. Some of 0:36:50.670 --> 0:36:52.589 the students don't want to use it. You know, they 0:36:52.640 --> 0:36:55.060 don't want to treat it like it's an inevitability and 0:36:55.080 --> 0:36:57.200 just like give in to the companies. But at the 0:36:57.239 --> 0:37:00.540 same time, they're trying to prepare their students, you know, 0:37:00.620 --> 0:37:03.640 for the real world and deal with their own students' 0:37:03.700 --> 0:37:07.779 anxieties about the job market. So it was just like, yeah, 0:37:07.800 --> 0:37:10.630 it was really fascinating. We had at the Washington Post, 0:37:10.850 --> 0:37:14.390 our like AI community, session about how to use it 0:37:14.450 --> 0:37:17.410 in journalism. I mean, everyone is struggling with this, right? 0:37:17.489 --> 0:37:19.910 And it just seems like teachers are maybe a little 0:37:19.930 --> 0:37:23.050 bit ahead of us and they're talking about having more transparency, 0:37:23.430 --> 0:37:25.730 having the students, like, if you don't punish them and 0:37:25.770 --> 0:37:28.319 you just talk to them about how they're using the tools, 0:37:28.780 --> 0:37:32.080 having those kind of like deeper discussions and AI literacy, 0:37:32.820 --> 0:37:34.880 I would say that teachers are doing better than the 0:37:34.900 --> 0:37:36.080 Washington Post at the moment. 0:37:37.120 --> 0:37:37.800 It's very tough. 0:37:38.540 --> 0:37:40.920 Did you see my thing I did last week, Natasha, 0:37:41.200 --> 0:37:45.620 on the AI writing? No, I missed it. There was 0:37:45.640 --> 0:37:49.880 this whole debate around because the Wall Street Journal editorial. 0:37:49.920 --> 0:37:50.640 Oh, yeah, yeah, yeah. 0:37:50.700 --> 0:37:54.070 Paul Jago was like the opinion editor at the journal 0:37:54.110 --> 0:37:57.350 was like, oh, who cares? Like, yeah, Druckenmiller used AI 0:37:57.370 --> 0:37:59.870 to write his guest column. Who cares? So I did 0:37:59.910 --> 0:38:03.150 like an analysis. I ran for the past month. I 0:38:03.210 --> 0:38:06.779 ran every single guest article or guest essay in the 0:38:06.969 --> 0:38:09.300 in the WSJ, the New York Times and the Washington 0:38:09.320 --> 0:38:12.569 Post through Pan Graham. and just to see like what 0:38:12.610 --> 0:38:16.670 percentage used AI. And it was like, it's pretty interesting. Like, like, 0:38:16.950 --> 0:38:19.150 you know, the, the, the New York times, which has 0:38:19.190 --> 0:38:21.650 is very anti AI had like, uh, yeah, I think 0:38:21.670 --> 0:38:25.350 it was like 11, um, columns that used AI that 0:38:25.390 --> 0:38:28.090 gas columns in the Washington post had, you know, a 0:38:28.160 --> 0:38:31.879 significant amount more than that. I forget the exact percentage. Um, 0:38:32.080 --> 0:38:33.839 but one of the, one of the ones that showed 0:38:33.880 --> 0:38:38.020 up as a hundred percent AI turned out to be, uh, the, the, um, 0:38:39.080 --> 0:38:42.109 I feel, I feel bad, like, call calling people out 0:38:42.150 --> 0:38:45.609 here but like you know it was a dartmouth um 0:38:46.030 --> 0:38:51.160 provost actually who who was writing about policing ai in school. 0:38:51.000 --> 0:38:54.100 Oh no stop it and i'm like. 0:38:53.860 --> 0:38:56.000 You guys like you can't it's to your point it's 0:38:56.020 --> 0:38:57.960 actually like come on you're not going you're not putting 0:38:57.980 --> 0:39:00.020 the genie back in the bottle here like kids are 0:39:00.040 --> 0:39:01.560 going to be using ai if you give them an 0:39:01.640 --> 0:39:03.719 essay to do outside of school they are for sure 0:39:03.739 --> 0:39:06.120 going to use ai you have to judge the end 0:39:06.160 --> 0:39:08.300 product right like just is it good. 0:39:08.800 --> 0:39:11.209 No, no, no. Well, that's not what the teachers are saying. 0:39:11.250 --> 0:39:13.850 They're saying we have to move away from that because 0:39:13.870 --> 0:39:18.529 that's not working. But some of them are even incorporating 0:39:18.590 --> 0:39:21.830 AI to do this. Like, you know, for a rhetoric class, 0:39:21.870 --> 0:39:24.090 like you have to have an argument with the chatbot 0:39:24.110 --> 0:39:27.130 that the teacher has designed, but the teacher then gets 0:39:27.810 --> 0:39:31.030 a record of your critical thinking, what questions are you asking? 0:39:31.090 --> 0:39:34.719 I mean, it's more work for absolutely everyone, like the teachers, 0:39:34.820 --> 0:39:38.259 the students. But I think they're saying you cannot look 0:39:38.320 --> 0:39:41.259 at the, you can't just like have them turn in 0:39:41.340 --> 0:39:42.520 something at the end. 0:39:42.580 --> 0:39:46.180 Measuring output is irrelevant. I mean, everyone can generate perfect 0:39:46.239 --> 0:39:46.719 output now. 0:39:47.020 --> 0:39:47.219 Yeah. 0:39:47.239 --> 0:39:47.319 Yeah. 0:39:47.570 --> 0:39:50.170 Yeah, it's a lot more like process oriented. And I 0:39:50.190 --> 0:39:53.489 think also they're just realizing you can't there's no technical solution. 0:39:53.950 --> 0:39:56.950 So you have to find like policies and ways of 0:39:57.030 --> 0:40:00.870 talking about it, you know, and just like keeping closer 0:40:00.930 --> 0:40:02.730 tabs on them as they're learning. 0:40:02.910 --> 0:40:05.950 I think the classroom should just be analog. That's my like, 0:40:06.050 --> 0:40:08.370 I just think you just don't have technology in the classroom. 0:40:08.430 --> 0:40:10.610 Well, that's what Zoran is trying to do right in 0:40:10.650 --> 0:40:13.250 New York City. And, you know, you saw the reaction 0:40:13.270 --> 0:40:14.910 to that. I don't know. 0:40:15.070 --> 0:40:17.010 But Taylor, sorry, just one point. 0:40:17.070 --> 0:40:19.509 This is the mayor of New York who's banned elementary 0:40:19.550 --> 0:40:22.690 school and middle school AI for the next year. 0:40:23.270 --> 0:40:27.510 He's banned AI, but that's stupid. I'm saying ban technology. 0:40:27.860 --> 0:40:28.340 Ban technology. 0:40:28.360 --> 0:40:28.880 It's for one year. 0:40:28.900 --> 0:40:32.239 Like, don't ban AI. Just say no screens. 0:40:32.420 --> 0:40:33.640 He said generative AI. 0:40:34.340 --> 0:40:36.620 Yeah, generative AI, but that's like too narrow. It's like, 0:40:36.800 --> 0:40:39.069 that's not the problem. The problem is like, once you 0:40:39.090 --> 0:40:41.480 put technology in the classroom, it just becomes like teachers 0:40:41.540 --> 0:40:43.839 are like, here's your iPad. Like this happens in my 0:40:43.880 --> 0:40:45.080 kid's school. And the parents are. 0:40:45.100 --> 0:40:46.859 The kids are only in the classroom for a small 0:40:46.900 --> 0:40:49.380 number of hours in a day relative to their total 0:40:49.940 --> 0:40:50.600 living hours. 0:40:50.700 --> 0:40:51.739 So... Yeah, this is. 0:40:51.719 --> 0:40:54.760 The problem, I think, with our whole public education system 0:40:54.820 --> 0:40:57.470 is that it's underfunded. These teachers are... These teachers, by 0:40:57.510 --> 0:41:00.569 the way, a lot of them are... They're overwhelmed. They 0:41:00.590 --> 0:41:02.910 have too many kids. Like, I mean, I've talked to 0:41:02.950 --> 0:41:04.890 teachers too. And it's like, I understand why they give 0:41:04.910 --> 0:41:06.569 the kids the iPad or whatever. They rely on these 0:41:06.590 --> 0:41:09.180 tools because they don't have the capacity to like, teach 0:41:09.239 --> 0:41:11.379 at the scale that they're being required to. And so 0:41:11.460 --> 0:41:14.480 I think we need a revamp of our entire sort 0:41:14.500 --> 0:41:17.419 of model of teaching so that it keeps up with 0:41:17.520 --> 0:41:20.080 modern times. We need to look at new ways of learning. 0:41:20.100 --> 0:41:22.080 I mean, we haven't sort of rethought the way that 0:41:22.100 --> 0:41:25.180 we teach in like a hundred years, you know? And 0:41:25.239 --> 0:41:27.299 so I think all of this new technology is coming 0:41:27.340 --> 0:41:30.700 so fast and we just need to, you know, whether 0:41:30.780 --> 0:41:33.719 it is, yeah, there's more offline time during the day, 0:41:33.739 --> 0:41:35.660 maybe like the first half of school is offline. The 0:41:38.660 --> 0:41:41.330 But surely we need to think in a more innovative way. 0:41:41.989 --> 0:41:44.469 Kids don't need to learn technology in school. They are 0:41:44.530 --> 0:41:47.210 going to learn the technology outside of school. 0:41:47.270 --> 0:41:51.220 I totally disagree because, Reid, if kids don't learn about 0:41:51.239 --> 0:41:53.900 technology in school, they don't learn literacy. And you're right 0:41:53.920 --> 0:41:56.160 that they will learn about it outside of school, but 0:41:56.180 --> 0:41:59.219 they don't learn those crucial literacy skills. And that's what 0:41:59.260 --> 0:42:00.799 I think we need to start teaching kids. 0:42:00.820 --> 0:42:02.920 But you don't need to have the technology in the 0:42:02.980 --> 0:42:05.220 school to teach them that kind of stuff, right? Because 0:42:05.239 --> 0:42:06.170 that's critical thinking. 0:42:07.640 --> 0:42:10.439 People are moving away from that too, right? Taking away 0:42:10.500 --> 0:42:14.940 cell phones in school. I think now that like Google has, 0:42:15.360 --> 0:42:18.259 I think a lot of parents were shocked by Google 0:42:18.300 --> 0:42:21.580 turning on, you know, because everyone has Chromebooks, Google turning 0:42:21.680 --> 0:42:25.160 on Gemini without like a lot of warning or a 0:42:25.200 --> 0:42:27.839 lot of input from other people. So I think they 0:42:27.980 --> 0:42:30.820 are rethinking that, but at least the teachers that I 0:42:30.840 --> 0:42:33.910 was talking to were also folks who were like, teach 0:42:33.950 --> 0:42:39.660 like continuing education or like people who are already working 0:42:39.700 --> 0:42:42.440 in the working world and they, you know, are taking 0:42:42.460 --> 0:42:44.819 classes and they just don't have, not everybody has the 0:42:44.960 --> 0:42:46.819 option to go back to blue books. And I think 0:42:46.930 --> 0:42:48.629 also a lot of people that I talked to just 0:42:48.650 --> 0:42:52.770 felt like that's not a solution. Also, everyone's handwriting is terrible. 0:42:52.930 --> 0:42:55.650 Well, that's, that's true. But the social dynamics question is interesting. 0:42:55.670 --> 0:42:59.089 Remember my history class at elementary school, the teacher would 0:42:59.210 --> 0:43:02.440 write an essay on the whiteboard And then we would 0:43:02.500 --> 0:43:05.160 copy the essay from the whiteboard into our notebooks. And 0:43:05.180 --> 0:43:07.880 then we would be graded on our ability to reproduce 0:43:07.940 --> 0:43:11.440 exactly what he'd written verbatim the next day. And so 0:43:11.480 --> 0:43:13.800 maybe I'm happy that that style of teaching is being 0:43:14.180 --> 0:43:17.210 replaced by technology. But I mean, we thought it was 0:43:17.230 --> 0:43:19.029 a little bit absurd when we were 12. But now 0:43:19.070 --> 0:43:21.609 you could just put what he'd written into AI and 0:43:21.730 --> 0:43:24.070 be told that it was lacking in all kinds of 0:43:24.130 --> 0:43:26.210 ways and come in the next day and say, you know, sir, 0:43:26.250 --> 0:43:28.340 what you told us yesterday is garbage. So I mean, 0:43:28.360 --> 0:43:32.580 there's an interesting total recalibration of the power dynamics between 0:43:32.600 --> 0:43:35.640 students and teachers when you have a reference point to 0:43:36.280 --> 0:43:38.020 all of human knowledge at the stroke of a. 0:43:38.000 --> 0:43:41.480 Keyboard, right? So it's kind of an interesting thing. Yeah, 0:43:41.620 --> 0:43:43.500 I think there's no reason to have a screen in 0:43:43.520 --> 0:43:46.940 the classroom. I think that's the... I mean, if there 0:43:46.960 --> 0:43:48.960 were research that showed that, oh, kids are going to 0:43:49.000 --> 0:43:51.259 learn better because... I think most of the research, though, 0:43:51.300 --> 0:43:52.580 shows the opposite. And it's like. 0:43:53.100 --> 0:43:54.280 No, that's not true. 0:43:54.500 --> 0:43:57.450 It doesn't. That's not... Taylor, like... I have conversations with 0:43:57.510 --> 0:43:58.690 my kids all the time. 0:43:58.790 --> 0:44:01.190 But Reed, you have to understand, and I think this 0:44:01.250 --> 0:44:03.370 is part of the problem, and I've been reporting a 0:44:03.410 --> 0:44:05.029 lot on these cell phone bans, which is a good 0:44:05.350 --> 0:44:08.910 example of this. It is so school dependent. We have 0:44:08.989 --> 0:44:11.800 such a wild, you know, an inner city school in 0:44:11.820 --> 0:44:14.400 Los Angeles has a completely different student body that has 0:44:14.440 --> 0:44:17.360 completely different needs and completely different teaching structure than a 0:44:17.400 --> 0:44:21.400 suburban school in Missouri. We need to have more tailored solutions. 0:44:21.460 --> 0:44:23.420 And I agree, I'm not a screen, I don't want 0:44:23.440 --> 0:44:25.250 these kids on screens in schools either, by the way. 0:44:25.270 --> 0:44:27.029 I agree with you on that, Reid. But I think 0:44:27.070 --> 0:44:29.969 we need to figure out, make sure that those kids 0:44:30.010 --> 0:44:32.350 aren't immediately left behind. Because when you take away all 0:44:32.370 --> 0:44:37.140 of these assistive tools, often, the result is they didn't 0:44:37.280 --> 0:44:41.259 fund more teachers that can teach people by hand. You 0:44:41.280 --> 0:44:43.500 know what I mean? The result is that kids are 0:44:43.780 --> 0:44:44.600 getting less educated. 0:44:44.800 --> 0:44:46.680 But what do you mean left behind? You mean like 0:44:46.719 --> 0:44:50.000 they don't have technology at home? No. Well, yes. 0:44:50.080 --> 0:44:51.960 A lot of them don't have technology at home. They 0:44:52.000 --> 0:44:55.020 don't have access to it. Their parents are incredibly technically 0:44:55.140 --> 0:44:57.350 illiterate or they're working double shifts and they don't have 0:44:57.390 --> 0:44:59.770 time to explain this stuff to their kids. Their kids 0:44:59.830 --> 0:45:02.710 end up in really bad places online. We need, I 0:45:03.930 --> 0:45:06.070 do think that we need more literacy. 0:45:06.410 --> 0:45:08.509 But you can teach literacy. I talk to my kids 0:45:08.650 --> 0:45:11.190 all the time about social media and all this, like 0:45:11.230 --> 0:45:13.210 the ills, the problems with what they need to look 0:45:13.270 --> 0:45:15.490 out for online if they're playing a video game or whatever. 0:45:15.510 --> 0:45:17.880 But Reid, you are a parent that is involved. A 0:45:17.920 --> 0:45:19.060 lot of these parents are not. 0:45:19.180 --> 0:45:20.919 Yeah, you are. No, that's not my point. My point 0:45:20.980 --> 0:45:24.080 is that teachers can have these conversations without using screens. 0:45:24.200 --> 0:45:26.460 Like I don't use a screen to talk to my 0:45:26.500 --> 0:45:28.859 kid about the dangers of Roblox. I just talk to them. 0:45:29.100 --> 0:45:31.950 That is because your kid is already immersed in this 0:45:31.989 --> 0:45:34.650 world and has these opportunities. I just, the thing that 0:45:34.710 --> 0:45:36.430 people need to realize when we're writing these laws is 0:45:36.469 --> 0:45:39.590 that there are kids with such a wide range of experiences. 0:45:39.920 --> 0:45:42.250 And I don't think bans or anything make a lot 0:45:42.270 --> 0:45:43.750 of sense. I think we need to look at each 0:45:43.850 --> 0:45:45.890 school and say, what are the problems with each of 0:45:45.930 --> 0:45:49.779 these schools? And instead of spending $ 5 million, as some 0:45:49.820 --> 0:45:53.400 school districts have, $ 5 million on yonder pouches to make 0:45:53.420 --> 0:45:56.570 sure the kids don't text their friends, we could invest in. 0:45:58.900 --> 0:46:02.549 In teachers and help, right? And social workers and all 0:46:02.590 --> 0:46:04.550 this stuff. I'm not kidding you. A school district literally 0:46:04.570 --> 0:46:06.549 has spent $ 5 million on this. No, I believe it. 0:46:06.550 --> 0:46:09.650 So I just think like there are smarter ways to 0:46:09.670 --> 0:46:10.430 go about this. 0:46:10.810 --> 0:46:14.890 They shouldn't have cell phones like it's just a distraction. 0:46:14.960 --> 0:46:16.680 And also like I'm not going to go into detail 0:46:16.700 --> 0:46:19.000 about this, but like even just like smart watches, which 0:46:19.040 --> 0:46:22.520 are popular with my in my circle to like, you know, 0:46:22.640 --> 0:46:24.719 keep track of your kids when they're whatever biking to 0:46:24.760 --> 0:46:27.859 school or something. like those got kids into trouble. Just 0:46:27.910 --> 0:46:28.750 a smartwatch. 0:46:28.790 --> 0:46:31.370 I mean, they- And so did beepers. And so did 0:46:31.489 --> 0:46:34.870 all technology. We cannot, we need our children to grow 0:46:34.989 --> 0:46:37.549 up in a world where they can navigate that world. 0:46:37.630 --> 0:46:40.210 And the most privileged kids are going to be totally 0:46:40.270 --> 0:46:42.600 fine because the most privileged kids have an enormous amount 0:46:42.620 --> 0:46:45.219 of support and learning outside of school. But a lot 0:46:45.239 --> 0:46:47.980 of these kids in underprivileged school districts are struggling. The 0:46:48.020 --> 0:46:50.260 teachers are struggling and we need to invest in education. 0:46:50.360 --> 0:46:53.300 And I think we need to rethink our public education 0:46:53.320 --> 0:46:55.339 system from the ground up in terms of how we teach. 0:46:55.440 --> 0:46:58.219 And maybe that is a lot of offline education, you know, 0:46:58.500 --> 0:47:01.620 but I think it also needs to include some level 0:47:01.760 --> 0:47:04.069 of media literacy, which a lot of these teachers don't have. 0:47:04.270 --> 0:47:08.350 Privileged kids have all kinds of problems with technology as well. 0:47:08.630 --> 0:47:12.410 I think this is actually a problem that spans socioeconomics. 0:47:12.770 --> 0:47:15.509 Yeah, but privileged kids are more likely to understand how 0:47:15.530 --> 0:47:19.060 to use these tools to get jobs, navigate these information ecosystems. 0:47:19.840 --> 0:47:22.790 That's all we have time for today. Thank you all 0:47:23.270 --> 0:47:23.960 so much for joining us. 0:47:23.989 --> 0:47:24.610 Oz is like, shut up. 0:47:27.350 --> 0:47:29.529 Well, we've solved government. We've solved education. 0:47:29.550 --> 0:47:31.270 Yes. Exactly. Yes. 0:47:31.790 --> 0:47:32.029 Yes. 0:47:32.050 --> 0:47:41.090 I think we put the world to right. For Tech Stuff, 0:47:41.390 --> 0:47:44.470 I'm Oz Voloshin. This episode was produced by Eliza Dennis. 0:47:45.130 --> 0:47:47.810 It was executive produced by me and Julia Nutter for 0:47:47.850 --> 0:47:52.480 Kaleidoscope and Katrina Norvell for iHeart Podcasts. Jack Inslee mixed 0:47:52.520 --> 0:47:55.550 this episode and Kyle Murdoch wrote our theme song. A 0:47:55.590 --> 0:47:59.489 special thank you to Taylor Lorenz, Natasha Tiku and Reid Albigotti. 0:47:59.950 --> 0:48:01.989 Please check out all the work they put out into 0:48:02.010 --> 0:48:04.490 the world. We're lucky to call them friends of the pod.