WEBVTT - Chinese Humanoid Robots Aren’t Welcome Here, US Says - Week in Tech 0:00:11.840 --> 0:00:13.920 Read, Taylor, I want to ask you to open this 0:00:13.920 --> 0:00:17.640 week's episode. Would you shop online if you didn't actually 0:00:17.720 --> 0:00:21.200 get anything in the mail. But I'm asking because I've 0:00:21.200 --> 0:00:24.320 recently read about these dopamine sites out of South Korea 0:00:24.400 --> 0:00:27.160 that let you look through non existent products with made 0:00:27.200 --> 0:00:29.320 up reviews and promotions that you can add to your cart, 0:00:29.480 --> 0:00:33.280 click buy, track your packages and never receive a thing, 0:00:33.360 --> 0:00:36.600 No money spent, just the satisfaction of buying something online. 0:00:37.159 --> 0:00:39.960 No you got, Taylor. 0:00:40.280 --> 0:00:43.040 I think I already do this with Zillo every night 0:00:43.120 --> 0:00:47.080 where I'm you know, perusing multimillion dollar houses in Pasadena 0:00:47.159 --> 0:00:48.800 adding them to my favorites. 0:00:49.080 --> 0:00:52.240 Like I don't do the Zilla thing because there's nothing 0:00:52.280 --> 0:00:54.920 on the market here in the day area, but I 0:00:55.320 --> 0:00:57.640 would know. Like what I would do, though, is could 0:00:57.680 --> 0:01:00.600 I order stuff but not have the package just show 0:01:00.680 --> 0:01:02.520 up at my front door, so that, like all my 0:01:02.560 --> 0:01:05.160 neighbors know that we're just like those horrible people who 0:01:05.319 --> 0:01:06.959 order like twenty things on Amazon every. 0:01:06.920 --> 0:01:09.520 Day that the note packages thing. 0:01:09.600 --> 0:01:12.080 For sure, I still want the product stuff. 0:01:16.840 --> 0:01:17.880 Welcome to Tech Stuff. 0:01:17.920 --> 0:01:20.400 I'm as Volocian and this is the week in Tech 0:01:20.720 --> 0:01:23.640 where I'm joined by the world's most clubs and reporters. 0:01:23.080 --> 0:01:25.560 To break down what's really happening in tech right now. 0:01:26.080 --> 0:01:28.960 Today, we're joined by Taylor Lorenzo use a mag and 0:01:29.080 --> 0:01:31.399 Read Albergotti, tech editor at Centophore. 0:01:31.440 --> 0:01:34.040 Welcome both, great to be here, Thanks for having us. 0:01:34.040 --> 0:01:37.040 Great to have you both, As always read, the FCC 0:01:37.240 --> 0:01:41.280 moved to ban Chinese humanoid robots and quadrupeds from entering 0:01:41.280 --> 0:01:45.440 the US, which I can't read without laughing. But I mean, 0:01:46.000 --> 0:01:47.319 what's what's going on here? 0:01:47.440 --> 0:01:47.520 Like? 0:01:47.600 --> 0:01:48.760 How serious is this story? 0:01:48.800 --> 0:01:50.480 What what does it have to do with US robot 0:01:50.520 --> 0:01:52.000 makers lobbying efforts? 0:01:52.040 --> 0:01:53.200 And what's gonna happen next? 0:01:53.880 --> 0:01:56.680 Yeah, I mean this is this is another sort of 0:01:56.720 --> 0:02:01.000 it's its industrial policy protectionism. I guess we shouldn't be surprised. 0:02:01.000 --> 0:02:04.240 I mean, they're already are no buy D Chinese cars 0:02:04.320 --> 0:02:07.360 in the US and it and on its face, you're 0:02:07.480 --> 0:02:10.239 you're you'd sort of put into that box and you say, okay, 0:02:10.280 --> 0:02:14.120 this this makes sense. We want US robot makers to 0:02:14.120 --> 0:02:16.280 to kind of have a clear run at this. But 0:02:16.360 --> 0:02:19.240 I think there's a there and and granted like there 0:02:19.280 --> 0:02:21.480 will be exceptions to this, right, this is not like 0:02:21.560 --> 0:02:23.680 this is not a law. This is you know, this 0:02:23.760 --> 0:02:27.040 is this is policy coming out of the White House essentially, 0:02:27.960 --> 0:02:30.240 but like on its face, you there are these big 0:02:30.320 --> 0:02:33.400 questions like a bunch if you go into any sort 0:02:33.400 --> 0:02:36.400 of like AI company that's doing anything in the physical 0:02:36.400 --> 0:02:38.919 world and in the Bay Area right now, they've got 0:02:38.960 --> 0:02:42.440 these Chinese robots like sitting around from all these different companies, 0:02:42.440 --> 0:02:44.560 and they're you know, some of them are cheap, like 0:02:44.639 --> 0:02:48.120 ten thousand apiece, and they're doing all sorts of testing 0:02:48.160 --> 0:02:51.320 on these things, right They're they're essentially building software and 0:02:51.400 --> 0:02:54.360 the and the hardware is what they use to you know, 0:02:54.440 --> 0:02:57.360 to gather data or just test out their their algorithm. 0:02:57.440 --> 0:03:02.720 So I think it's actually like potentially detrimental to innovation. 0:03:02.919 --> 0:03:04.600 So this is all and it's the same thing with 0:03:04.680 --> 0:03:07.760 like the same issue with these Chinese open source models. 0:03:07.800 --> 0:03:09.680 Right It's like, I want to talk about the open 0:03:09.680 --> 0:03:11.720 source models, but let's stay with the robots for a 0:03:11.720 --> 0:03:14.239 little bit. I mean, the detail about the quadrupeds was 0:03:14.280 --> 0:03:16.160 the one that really really caught my eye because I 0:03:16.200 --> 0:03:18.440 kind of thought the four legged robots were only for 0:03:18.480 --> 0:03:21.920 YouTube videos that I didn't realized they were threat to America. 0:03:22.080 --> 0:03:25.040 I saw I was in Boston a couple of weeks 0:03:25.040 --> 0:03:27.880 ago and I saw some people walking their their robot 0:03:28.000 --> 0:03:30.359 dicers that looked like, yeah, they looked like they were 0:03:30.440 --> 0:03:33.960 sort of like MIT researchers or something. But no, I 0:03:33.960 --> 0:03:37.280 mean it's it's like the same issue though. It's the 0:03:37.280 --> 0:03:39.920 same principle, which is like you try to you try 0:03:39.960 --> 0:03:42.400 to stop the you know, the Chinese from encroaching in 0:03:42.440 --> 0:03:45.280 the US market, and then you sort of like also 0:03:45.480 --> 0:03:48.680 hinder your own innovation. So these things always come at 0:03:48.680 --> 0:03:50.440 a cost. So I think that the trick is going 0:03:50.520 --> 0:03:52.360 to be figuring out how to thread that needle right, Like, 0:03:52.880 --> 0:03:55.000 do what you need to do, you know, if you're 0:03:55.040 --> 0:03:57.720 the White House, to to sort of protect US companies, 0:03:58.040 --> 0:04:00.760 stop the Chinese from i don't know, dumping robots if 0:04:00.760 --> 0:04:05.160 that's even a thing, but then also allow US researchers 0:04:05.240 --> 0:04:08.600 to still, you know, to still be able to innovate 0:04:08.720 --> 0:04:10.760 and come. You know, you don't want like a situation 0:04:10.800 --> 0:04:13.800 where there's like one or two robotics companies in the 0:04:13.920 --> 0:04:16.200 US and if they don't, if they don't just happen 0:04:16.240 --> 0:04:18.880 to win this race outright, you know, the US is 0:04:19.560 --> 0:04:20.279 way behind. 0:04:20.480 --> 0:04:22.400 This will be like the cause in the Soviet Union 0:04:22.440 --> 0:04:24.720 when the when in nineteen eighteen nine, Right, the US 0:04:24.760 --> 0:04:28.719 will have all of these like Dusty Treban style robots 0:04:28.800 --> 0:04:32.039 call the Tita. Have you been following the robot element 0:04:32.080 --> 0:04:32.919 of this story. 0:04:32.880 --> 0:04:36.040 Yes, I've been following. And does this mean that rumbas 0:04:36.080 --> 0:04:40.000 are banned? I saw some headline basically implying. 0:04:39.600 --> 0:04:42.839 That I have a Chinese I have a Chinese robot vacuum. 0:04:42.839 --> 0:04:47.880 That's really really good. I think it's a way outperforms 0:04:47.920 --> 0:04:52.520 the rumba. So I hope, I hope that's not the case. Look, 0:04:53.040 --> 0:04:55.599 I think this is like all of these policies. It's 0:04:55.640 --> 0:04:58.039 just the thing to remember is that you can do 0:04:58.120 --> 0:05:00.560 whatever you want to try to stop China these tech 0:05:00.640 --> 0:05:03.360 and you know, stop them from i don't know, catching 0:05:03.440 --> 0:05:07.640 up or ripping off American technology. The US will not 0:05:07.839 --> 0:05:11.160 win unless it can move faster than China on innovation. 0:05:11.560 --> 0:05:14.039 That is the most important thing about this race, and 0:05:14.040 --> 0:05:16.640 I think it often. I think it often gets sort 0:05:16.680 --> 0:05:18.320 of forgotten or pushed to the side. 0:05:18.720 --> 0:05:22.039 I totally agree with Red and yeah, I mean I 0:05:22.080 --> 0:05:25.200 was just reading this article on the Verge talking about 0:05:25.240 --> 0:05:27.479 like what's actually sort of included in this span and 0:05:27.520 --> 0:05:31.120 it's significantly broader than you would think. I mean, Obviously 0:05:31.200 --> 0:05:33.520 we're talking about the humanoid robots and the quadrupeds, but 0:05:33.560 --> 0:05:36.720 it's not just that, it's it's far broader. This ban 0:05:36.880 --> 0:05:40.839 covers basically any new software controlled robot that travels over 0:05:40.880 --> 0:05:44.000 the ground or weighs more than four point four pounds, 0:05:44.000 --> 0:05:47.360 including the DOC, and that can perceive its own environment 0:05:47.480 --> 0:05:50.880 and has wireless which is basically any robot vacuum like 0:05:51.160 --> 0:05:54.320 it's it's a lot of like household devices. Like I 0:05:54.320 --> 0:05:58.120 think people don't realize as well just how much stuff 0:05:58.160 --> 0:06:01.320 comes from China, Like how like automated I have this 0:06:01.400 --> 0:06:04.560 like automated you know, like watering system thing that I 0:06:04.600 --> 0:06:07.640 got from Amazon. I'm sure it's some Chinese little robot thing. 0:06:07.760 --> 0:06:08.120 Oh you did. 0:06:08.160 --> 0:06:10.120 We were talking about the garden last week and you 0:06:10.160 --> 0:06:12.240 were talking about wanting a gunting robot. And now you've 0:06:12.320 --> 0:06:13.560 now you filled the trigger. 0:06:13.960 --> 0:06:19.039 Exactly and I've ordered a Chinese gardening you know, watering thing. 0:06:19.160 --> 0:06:22.599 Probably that's potentially going to be illegal, legal, but but 0:06:22.760 --> 0:06:25.040 I know I snuck in right under the van. I 0:06:25.520 --> 0:06:27.760 just think, like, I mean, what Reid said is true. 0:06:27.960 --> 0:06:30.840 I think this is ridiculous. Like I understand from a 0:06:30.880 --> 0:06:33.719 consumer perspective, you know, with the electric cars. Okay, protecting 0:06:33.760 --> 0:06:36.760 our car industry in some way, but you know, at 0:06:36.839 --> 0:06:42.719 some point, robotics are becoming so pervasive, and this just 0:06:42.760 --> 0:06:45.520 seems like, you know, we we need to compete harder, 0:06:45.680 --> 0:06:48.480 like we can't. We're I feel like we're the declining 0:06:48.520 --> 0:06:51.679 and we're just trying to like protect this declining industry 0:06:51.680 --> 0:06:53.919 and we need to allow our researchers to compete. And 0:06:53.920 --> 0:06:56.680 and that means that you know, giving technologists access to 0:06:56.720 --> 0:06:57.440 this cutting. 0:06:57.160 --> 0:07:00.080 Head technology on the kind of robotspan Is this like 0:07:00.279 --> 0:07:01.960 kind of a thing that people expect to be in 0:07:02.040 --> 0:07:04.279 forced or is this kind of like a staking of 0:07:04.320 --> 0:07:05.640 some political kind of pulturing. 0:07:05.640 --> 0:07:07.760 Do we do we have any idea? 0:07:07.839 --> 0:07:10.040 Well, I like I was sort of getting at earlier. 0:07:10.240 --> 0:07:13.000 I mean, this is this is just you know, these 0:07:13.040 --> 0:07:15.320 are just orders coming out of the White House. So 0:07:15.480 --> 0:07:17.800 it's not like this is not a law. And I 0:07:17.800 --> 0:07:21.400 think they always they you know, like I think at 0:07:21.400 --> 0:07:23.240 some point they're going to go, oh my god, like 0:07:23.760 --> 0:07:25.640 this is way too broad. We got to find loop. 0:07:25.760 --> 0:07:28.560 You know, there's gonna be loopholes or or carve outs 0:07:28.560 --> 0:07:32.200 for specific areas, like you know, this is what always happens. 0:07:32.200 --> 0:07:34.720 It's like this this shoot from the hip sort of 0:07:35.080 --> 0:07:39.680 you know, proclamation policy, I guess strategy, and then and 0:07:39.720 --> 0:07:41.240 then you go back and make all it. And the 0:07:41.280 --> 0:07:43.680 same thing happened with tariffs, right It's like, oh my god, 0:07:43.680 --> 0:07:45.160 we have all these tariffs. And then it's like, well, 0:07:45.200 --> 0:07:47.480 this thing's carved out and that thing's carved out, and 0:07:47.520 --> 0:07:50.480 you eventually get to like, I don't know someplace where 0:07:50.480 --> 0:07:52.760 it's where it's logical. So I just take all of 0:07:52.800 --> 0:07:54.400 it with a grain of salt and just figure it's 0:07:54.400 --> 0:07:55.280 all going to change. 0:07:55.720 --> 0:07:58.720 I just think in the meantime, the damage is done. 0:07:58.920 --> 0:08:01.080 I mean, the tariffs are good example, that was like 0:08:01.120 --> 0:08:05.360 a huge economic shock. It caused prices to increase for consumers, 0:08:05.400 --> 0:08:07.520 like it did cause a lot of I would argue 0:08:07.520 --> 0:08:10.040 harm to our economy. And I think when we do 0:08:10.120 --> 0:08:13.880 things like this as well, like yeah, maybe you're right 0:08:14.000 --> 0:08:16.080 read and I agree, like it will have to be 0:08:16.160 --> 0:08:18.680 rolled back in some way because it's nonsensical. But in 0:08:18.720 --> 0:08:21.080 the meantime, it's hurt innovation in the meantime, it slowed 0:08:21.160 --> 0:08:23.720 us down. In the meantime, it's created liability for startups 0:08:23.760 --> 0:08:26.600 that don't want to take on any liability. So it 0:08:26.640 --> 0:08:27.320 just seems bad. 0:08:27.560 --> 0:08:28.960 Yeah, I mean, it'd be one thing if they did 0:08:28.960 --> 0:08:30.880 this and they're like, and we're going to put you know, 0:08:31.800 --> 0:08:33.640 more money into academic research. 0:08:33.720 --> 0:08:34.679 That was my next question. 0:08:34.760 --> 0:08:37.040 So is there going to be is there any commensurate 0:08:37.120 --> 0:08:40.360 effort to stimulate the American robotics industry. I so Deep 0:08:40.400 --> 0:08:43.120 Mind had an announcement today about some some new kind 0:08:43.120 --> 0:08:45.959 of robotics program and obviously read you we talked about 0:08:46.240 --> 0:08:48.680 robotics with applied intuition last week on the show. So 0:08:48.720 --> 0:08:50.400 it seems like there is like some action in the 0:08:50.480 --> 0:08:52.760 American robotics sphere. 0:08:53.200 --> 0:08:55.760 Of course, this is a huge industry. There's tons of 0:08:55.800 --> 0:08:58.360 private money going into it. But I think if you're 0:08:58.440 --> 0:09:00.480 the US, like the private sectors going to do what 0:09:00.520 --> 0:09:03.800 it's going to do, the levers that they can pull, 0:09:04.040 --> 0:09:07.440 I think are an academic research. Right. That's where like 0:09:07.480 --> 0:09:09.960 if you think the robotics race is over, like I 0:09:10.000 --> 0:09:12.120 think that's insane. Like I think there's still a lot 0:09:12.160 --> 0:09:14.200 of innovation that's going to happen, and it's maybe not 0:09:14.240 --> 0:09:16.080 all going to come from the private sector. Like if 0:09:16.120 --> 0:09:18.720 you want to do something policy wise, like you know, 0:09:19.120 --> 0:09:22.600 push forward research, right, basic research that's going to you know, 0:09:22.679 --> 0:09:26.000 help advance this industry. There's so many different areas that 0:09:26.120 --> 0:09:29.000 would speed things up. It's it's long term thinking, though, 0:09:29.080 --> 0:09:31.800 right and I think that's what Washington is really short 0:09:31.800 --> 0:09:32.360 on right now. 0:09:32.559 --> 0:09:35.040 I mean, I also just think of like the drama 0:09:35.080 --> 0:09:37.560 over the H one B visas and the way that 0:09:37.600 --> 0:09:43.520 we're treating immigrants. And there are so many talented scientists, engineers, 0:09:43.640 --> 0:09:46.920 roboticists that you know, might be from India, might be 0:09:47.000 --> 0:09:50.480 from other places that we've also pushed out of our country, 0:09:50.480 --> 0:09:53.480 and I think, I mean, I know that's affected startup's 0:09:53.559 --> 0:09:56.680 ability to recruit. I was just talking to somebody last 0:09:56.760 --> 0:09:59.319 night who is living in Canada and what is has 0:09:59.360 --> 0:10:00.920 a job in a Mameric can start up but can't 0:10:00.960 --> 0:10:03.840 start because of this visa process. This is just it's 0:10:03.920 --> 0:10:04.920 really ridiculous. 0:10:05.600 --> 0:10:08.160 Yeah, I mean, it's a it's a this is like 0:10:08.240 --> 0:10:11.320 such a challenge because I mean, what you really want 0:10:11.360 --> 0:10:14.320 to do is like is not sort of bam the 0:10:14.400 --> 0:10:17.080 Chinese robots, but find the people who are really the 0:10:17.120 --> 0:10:20.120 most talented people in China building the robots, bring them 0:10:20.120 --> 0:10:23.240 to the US, have them build the robots here. But 0:10:23.360 --> 0:10:26.040 then there's this worry that they're all spies. Or they 0:10:26.040 --> 0:10:28.640 all will be spies, right, and and there is I mean, 0:10:28.679 --> 0:10:31.839 these are legitimate concerns, but I think there's like a 0:10:32.000 --> 0:10:34.439 I mean this happened in that you mentioned the Soviet 0:10:34.480 --> 0:10:36.360 era too. I mean this happened in the Soviet era 0:10:36.480 --> 0:10:38.800 like the US because of its system, because of its 0:10:38.880 --> 0:10:44.160 open system, you know. There it makes you know, spying easier, 0:10:44.240 --> 0:10:46.880 and so that's just a downside. But then that open 0:10:46.920 --> 0:10:50.480 system leads to like massive innovation, so it's always a 0:10:50.520 --> 0:10:51.000 trade off. 0:10:51.480 --> 0:10:53.920 I got of liked on that word open. Let's talk 0:10:53.920 --> 0:10:55.800 about open source models. I think for the third week 0:10:55.800 --> 0:10:58.679 in a row this is a textuff record, but the 0:10:58.720 --> 0:11:03.079 news keeps rumbling on. So another Chinese export, of course 0:11:03.280 --> 0:11:07.240 is open source models from Chinese companies like Zai and 0:11:07.320 --> 0:11:08.199 moonshot Ai. 0:11:09.720 --> 0:11:11.439 Why is it study in the news for so long? 0:11:11.520 --> 0:11:15.319 And also read you wrote about this concern about censorship 0:11:15.480 --> 0:11:18.760 an Ai models Chinese models that may be unfounded. 0:11:19.520 --> 0:11:22.559 Yeah, there was this research study I wrote about yesterday 0:11:23.040 --> 0:11:27.160 from CTGT which does this interpretability research, like looking at 0:11:27.240 --> 0:11:30.680 how these models actually work so they can build you know, 0:11:30.800 --> 0:11:33.400 products for the private sector that are sort of harnessed 0:11:33.440 --> 0:11:36.920 and don't go off the rails. They did this experiment 0:11:36.960 --> 0:11:40.480 where they took a deep seek model and they fine 0:11:40.520 --> 0:11:43.800 tuned from that model and then tested the before and 0:11:43.840 --> 0:11:47.960 after Chinese censorship and found that the censorship was essentially 0:11:48.000 --> 0:11:50.959 gone once they once they built this new model, the 0:11:51.280 --> 0:11:54.800 base of it was was an open ai OSS like 0:11:54.920 --> 0:11:57.719 open weights model, and then it was fine tuned or 0:11:57.800 --> 0:12:01.200 taught from deep Seek. And so there have been this 0:12:01.200 --> 0:12:04.880 this talk of like subliminal messages being passed on from 0:12:04.920 --> 0:12:09.160 the teacher model to the student model. It gets it 0:12:09.200 --> 0:12:11.679 gets pretty technical and you can get into weeds. But 0:12:11.760 --> 0:12:14.160 like the long story short is like we need more 0:12:14.200 --> 0:12:17.120 research on this. But what it looks like is I 0:12:17.160 --> 0:12:21.559 think these these Chinese models are actually like not as 0:12:21.640 --> 0:12:24.400 much of a propaganda risk as you think because they 0:12:24.440 --> 0:12:28.320 will get changed and filtered through all these different techniques. 0:12:28.920 --> 0:12:30.840 And it also goes to this research point. I mean, 0:12:30.880 --> 0:12:34.360 this is like really interesting research that wouldn't happen unless 0:12:34.400 --> 0:12:37.600 there were these Chinese open source models. So it's another 0:12:38.720 --> 0:12:42.319 I guess point in favor of the pro open source people. 0:12:42.960 --> 0:12:47.200 Yeah. Absolutely, I mean I just think that like it's 0:12:47.240 --> 0:12:50.880 it's really concerning to me that like our response is 0:12:50.920 --> 0:12:54.000 to ban things like I you know, I watched everything 0:12:54.040 --> 0:12:57.320 that happened with TikTok. I covered the musically sale to 0:12:57.360 --> 0:13:00.080 Bite dance, which the government was totally fine with, and 0:13:00.080 --> 0:13:02.280 then they turned around five years later and they were like, actually, 0:13:02.320 --> 0:13:06.200 we're gonna ban TikTok, you know, based on nothing really 0:13:06.280 --> 0:13:11.280 except that they it was keep challenging meta products, you know, 0:13:11.440 --> 0:13:13.600 like there were all these like sort of like made 0:13:13.679 --> 0:13:17.280 up concerns about China that never came to fruition. And 0:13:17.320 --> 0:13:19.640 similar to the Chinese open source models. I mean, we 0:13:19.720 --> 0:13:21.679 did have a lot of research on TikTok and there 0:13:21.760 --> 0:13:25.920 ultimately was just no evidence of anything that the government 0:13:26.040 --> 0:13:30.800 was claiming. And I just it's like if they had 0:13:30.920 --> 0:13:33.199 proof of these things and if they were doing these 0:13:33.240 --> 0:13:35.959 research and showing wow, the Chinese open source models are 0:13:36.080 --> 0:13:38.760 really like biased. So this could be really dangerous for 0:13:38.800 --> 0:13:41.560 American companies because if X Y z all right, then 0:13:41.600 --> 0:13:43.240 I think, you know, we could have a talk about it. 0:13:43.280 --> 0:13:46.040 But in this case, it just it's it seems like 0:13:46.120 --> 0:13:48.959 these speculative fears and as read mentioned earlier, it feels 0:13:49.080 --> 0:13:52.440 very cold war. It feels very like punching at ghosts, 0:13:52.480 --> 0:13:55.800 and I just I really worry about the damage that 0:13:55.840 --> 0:13:58.440 we're doing to our own tech landscape. 0:13:58.480 --> 0:14:02.120 In the meantime you mentioned, I think Markackerberg did two 0:14:02.160 --> 0:14:03.960 interviews this week, with one of the New York Times 0:14:03.960 --> 0:14:06.960 and one with the Journal, his first first media appearances 0:14:07.040 --> 0:14:10.560 since since his uh since his feed me guest lecture 0:14:10.640 --> 0:14:11.840 a few weeks ago. 0:14:12.720 --> 0:14:12.800 And. 0:14:14.360 --> 0:14:17.560 He was basically complaining about the dominance of of open 0:14:17.600 --> 0:14:18.599 a and ionanthropic. 0:14:18.720 --> 0:14:18.840 Right. 0:14:19.000 --> 0:14:21.040 Is that? Is that a fair summary of what the 0:14:21.080 --> 0:14:21.960 media tour was about. 0:14:22.480 --> 0:14:25.600 Yeah, I think there's two minds of this. I mean, one, 0:14:26.480 --> 0:14:29.000 you know, Meta or Facebook has always been sort of 0:14:29.000 --> 0:14:32.120 a pro open source company. They've developed a bunch of 0:14:32.320 --> 0:14:35.880 open source you know, protocols, and but on the other hand, 0:14:35.960 --> 0:14:38.400 I mean you have to sort of take this as like, yeah, 0:14:38.400 --> 0:14:40.240 I mean they're behind in the air raight, so like 0:14:40.400 --> 0:14:41.960 of course, I mean they're going to just be in 0:14:42.000 --> 0:14:45.400 favor of whatever slows their competitors down. So you know, 0:14:45.640 --> 0:14:46.080 there's that. 0:14:46.840 --> 0:14:48.360 There was a headline in the in the New York 0:14:48.360 --> 0:14:51.960 Times Silicon Valley splits over closing the borders to Chinese AI. 0:14:52.640 --> 0:14:54.560 So read is what you're saying that people who want 0:14:54.560 --> 0:14:56.360 to close the borders to Chinese AI are the ones 0:14:56.360 --> 0:14:58.120 who are currently winning and the ones who don't want 0:14:58.160 --> 0:14:59.920 to close the borders of Chinese AI, the ones who 0:15:00.120 --> 0:15:01.200 only losing a no note. 0:15:01.280 --> 0:15:03.920 To me, it's you know, that's a kind of misleading headline. 0:15:04.080 --> 0:15:07.320 To me, it feels very like Anthropic versus everyone else 0:15:07.440 --> 0:15:12.560 in tech. I mean, you had Google, like every single 0:15:12.600 --> 0:15:15.440 tech company kind of signed this open letter right against 0:15:15.440 --> 0:15:17.400 that sort of I felt like was a big subtweet 0:15:17.400 --> 0:15:20.240 of Anthropic. Dario comes out, He's like, Oh no, we 0:15:20.280 --> 0:15:22.160 don't like really want to be an open source models, 0:15:22.160 --> 0:15:23.600 but also we kind of think they're dangerous and we 0:15:24.000 --> 0:15:26.680 kind of I think like dark to me, like I 0:15:26.760 --> 0:15:31.200 see Anthropic running the meta playbook exactly where they have 0:15:31.320 --> 0:15:33.440 this dominance in the market right now and they're trying 0:15:33.480 --> 0:15:36.760 to sort of do regulatory capture to maintain it. 0:15:38.200 --> 0:15:40.120 More about the letter, Taylor, just before we go too far? 0:15:40.640 --> 0:15:44.880 Sure so that So Jensen Wang, CEO of Nvidia, joins 0:15:44.920 --> 0:15:47.040 Twitter big. By the way, I just have to say 0:15:47.080 --> 0:15:49.920 Mark Zuckerberg's on Twitter promoting his Wall Street Journal op 0:15:50.040 --> 0:15:52.600 ed like Jensen's on Twitter. I feel like Twitter has 0:15:52.640 --> 0:15:55.920 gotten peak lately. I love to see all these tech 0:15:55.960 --> 0:15:59.880 CEOs on there. But so Jensen joins Twitter and po 0:16:00.360 --> 0:16:03.120 this open letter, which obviously is you know, shared to 0:16:03.160 --> 0:16:07.280 the website whatever, and it is a defensive open source 0:16:07.600 --> 0:16:11.040 open weight models. Obviously Nvidia has a strong you know, 0:16:11.120 --> 0:16:14.440 business interest in this as well, but it becomes this letter, 0:16:14.520 --> 0:16:16.680 this open letter becomes this sort of like rallying cry, 0:16:16.800 --> 0:16:20.000 and you have just dozens and dozens of tech companies 0:16:20.000 --> 0:16:23.080 sign on to it, you know, by the by the weekend, 0:16:23.280 --> 0:16:25.840 like every almost every major tech company it'd signed on 0:16:25.840 --> 0:16:28.520 to it, including all these startup you know why, y Combinator, 0:16:28.520 --> 0:16:31.360 et cetera, a bunch of startups. So it felt like 0:16:32.200 --> 0:16:36.160 basically the entire tech industry had united to advocate for 0:16:36.400 --> 0:16:39.359 open weight AI models to be accept to remain accessible. 0:16:39.920 --> 0:16:43.920 And then Anthropic came out. Dario issued this this blog 0:16:43.960 --> 0:16:48.560 post saying basically like trying again doing the meta thing 0:16:48.600 --> 0:16:51.040 where it's like, oh, we don't really want to ban TikTok. 0:16:51.080 --> 0:16:53.080 We just think it's a national security thread and it 0:16:53.080 --> 0:16:55.600 should be eliminated and probably we shouldn't be able to 0:16:55.600 --> 0:16:57.960 download the United States and so like That's kind of 0:16:57.960 --> 0:16:59.960 how his letter read to me. I don't know what 0:17:00.040 --> 0:17:01.720 read thought, but it was like, oh no, no, no, we 0:17:02.080 --> 0:17:05.080 don't mind open Wait we just think it's incredibly dangerous 0:17:05.119 --> 0:17:07.040 and you know, I don't like to you know, that's 0:17:07.040 --> 0:17:08.359 probably not worth the resk or whatever. 0:17:08.440 --> 0:17:10.600 So read on Smiley, I want to hear your tape 0:17:10.600 --> 0:17:12.560 of first. Can one of you help us under sound 0:17:12.560 --> 0:17:14.200 difference between open source and open weight? 0:17:14.600 --> 0:17:16.879 Yeah, I mean it's just a this is just a 0:17:16.920 --> 0:17:21.399 semantics thing really because these models don't reveal their underlying 0:17:21.480 --> 0:17:24.640 data sets. And so people in the open source community 0:17:24.640 --> 0:17:26.640 took issue with that term because it's like, well, it's 0:17:26.640 --> 0:17:30.480 not totally open, but you know, you can still for 0:17:30.520 --> 0:17:34.199 all intents and purposes, they're open. I use yeah, I 0:17:34.320 --> 0:17:36.840 use open source all the time because I think it 0:17:37.000 --> 0:17:39.080 just confuses people to a lot of people to say 0:17:39.119 --> 0:17:41.840 open weights. But but you know, there's a there's a 0:17:42.000 --> 0:17:44.480 sort of debate there. There was a big debate about 0:17:44.480 --> 0:17:47.840 this in the open source community. But no, I mean, 0:17:47.920 --> 0:17:51.399 I basically agree with Taylor on this. I think the 0:17:51.480 --> 0:17:53.320 one thing I would sort of take issue with is 0:17:53.359 --> 0:17:56.760 like I actually think that anthropic people in Dario, like 0:17:56.800 --> 0:17:59.680 they're like true believers and they really are scared of 0:18:00.640 --> 0:18:03.119 like AI and what's going to happen if you know, 0:18:03.200 --> 0:18:05.760 you have these open models that can have all the 0:18:05.800 --> 0:18:09.800 guardrails and protection stripped away from them that eventually We're 0:18:09.800 --> 0:18:13.560 not there yet, but eventually, I think pose like a 0:18:13.640 --> 0:18:17.800 large amount of risk in cybersecurity or bioterrorism. 0:18:17.080 --> 0:18:20.159 Doesn't it make us safer? Like, isn't the argument, like 0:18:20.200 --> 0:18:22.879 the argument that I've seen from security people and Alex 0:18:22.960 --> 0:18:25.480 Damos was talking about this sort of a long time 0:18:25.480 --> 0:18:30.360 security guy at Metali. Actually it strengthens everyone to have 0:18:30.480 --> 0:18:33.040 these open source models out there and available because it 0:18:33.080 --> 0:18:35.240 allows I mean the Hugging Face incident is a perfect 0:18:35.240 --> 0:18:37.280 example of this, Like when you have these guard bills 0:18:37.320 --> 0:18:39.040 and you have these restrictions and we're only allowed to 0:18:39.119 --> 0:18:41.800 use anthropic robe and ai basically like it actually made 0:18:41.800 --> 0:18:43.400 it makes it more dangerous. 0:18:43.680 --> 0:18:48.320 Yeah, the hugging Face example was when open Aiy's cyber 0:18:48.440 --> 0:18:51.800 tool kind of went semi rogue in a non properly 0:18:51.920 --> 0:18:55.600 send boxed environment and hecked hugging Face and Hugging Face 0:18:55.640 --> 0:18:58.040 couldn't defend themselves within the US models because of the 0:18:58.040 --> 0:19:00.399 god rails, and so they used a chiny models to 0:19:00.400 --> 0:19:02.040 defend themselves, which we talked about last week. 0:19:02.160 --> 0:19:05.000 Yeah, I mean that's where I would also agree with you, 0:19:05.040 --> 0:19:08.000 like I just my only point was, like, my only 0:19:08.000 --> 0:19:10.400 point of disagreement with you was, like, I actually think 0:19:10.440 --> 0:19:13.840 it's not just some sort of cynical regulatory capture point. 0:19:13.920 --> 0:19:16.400 I think they actually believe in this stuff. I don't 0:19:16.400 --> 0:19:19.440 think their conclusion though, is always right. Like I agree 0:19:19.480 --> 0:19:23.040