WEBVTT - Are You Ready For A ‘Hot Surveillance Summer’? - Week in Tech

0:00:11.800 --> 0:00:14.680
<v Speaker 1>Natasha Taylor. There was a fascinating story and Wired last

0:00:14.680 --> 0:00:19.120
<v Speaker 1>week about an internal memo Meta regarding a loose squirrel.

0:00:19.320 --> 0:00:22.400
<v Speaker 2>I did see this, Yeah, I didn't understand.

0:00:22.600 --> 0:00:24.119
<v Speaker 3>We had a bird get into the New York Times

0:00:24.160 --> 0:00:26.360
<v Speaker 3>office once and it did derail our afternoon.

0:00:26.480 --> 0:00:28.360
<v Speaker 1>So this was this was a hard hard to pass.

0:00:28.400 --> 0:00:32.720
<v Speaker 1>But there was a squirrel mails to Meta in the Bangkok.

0:00:32.440 --> 0:00:33.959
<v Speaker 3>Office, the poor squirrel.

0:00:34.520 --> 0:00:37.360
<v Speaker 1>There was great hilarity because it's been a rough time

0:00:37.400 --> 0:00:40.239
<v Speaker 1>for Meta, but people were commenting on how this was

0:00:40.280 --> 0:00:42.800
<v Speaker 1>the most joyful moment the company had for a while,

0:00:42.840 --> 0:00:45.080
<v Speaker 1>despite the fact that Janet had got scratched on his finger.

0:00:45.640 --> 0:00:47.000
<v Speaker 3>Well, I hope he got a Rabi's shot.

0:00:50.159 --> 0:00:52.600
<v Speaker 1>Welcome to Tech Stuff. I'm was Vloscha and this is

0:00:52.640 --> 0:00:54.480
<v Speaker 1>the Week in Tech where I'm joined by the world's

0:00:54.520 --> 0:00:57.600
<v Speaker 1>most plugged in reporters to break down what's really happening

0:00:57.600 --> 0:01:00.200
<v Speaker 1>in tech right now. Today we're joined by Taylor, the

0:01:00.240 --> 0:01:03.040
<v Speaker 1>rends if use a mag and Natasha Tiku, tech reporter

0:01:03.160 --> 0:01:04.160
<v Speaker 1>at the Washington Post.

0:01:04.240 --> 0:01:07.679
<v Speaker 3>Welcome both, Hey everybody, Hello, thanks for having us.

0:01:08.040 --> 0:01:08.840
<v Speaker 1>It's good to see you.

0:01:09.400 --> 0:01:09.800
<v Speaker 3>Natasha.

0:01:09.880 --> 0:01:13.800
<v Speaker 1>You recently published an article titled Inside the Secret Ai

0:01:13.840 --> 0:01:18.280
<v Speaker 1>War between Silicon Valley and China. Great headline, but it

0:01:18.319 --> 0:01:21.880
<v Speaker 1>wasn't about ships or competing for talent, but actually about

0:01:21.880 --> 0:01:25.360
<v Speaker 1>the models themselves. Can you explain what's going on here?

0:01:25.680 --> 0:01:30.000
<v Speaker 4>So this is something of a narrative war between the

0:01:30.160 --> 0:01:34.560
<v Speaker 4>US leading AI companies and the leading Chinese AI companies.

0:01:35.160 --> 0:01:39.160
<v Speaker 4>So what companies like Anthropic and Open Ai are alleging

0:01:39.480 --> 0:01:44.400
<v Speaker 4>is that these Chinese companies are accessing their models which

0:01:44.440 --> 0:01:47.800
<v Speaker 4>are not supposed to be available in China and Hong Kong.

0:01:48.160 --> 0:01:51.680
<v Speaker 4>They're accessing their models in order to do this process

0:01:51.680 --> 0:01:55.840
<v Speaker 4>called distillation, which basically involves kind of you know, distilling,

0:01:55.960 --> 0:01:58.680
<v Speaker 4>taking the learnings from a larger model and applying it

0:01:58.720 --> 0:02:02.320
<v Speaker 4>to us to improve a smaller model. And distillation is

0:02:02.400 --> 0:02:05.120
<v Speaker 4>a process that has been used in the AI industry

0:02:05.200 --> 0:02:08.520
<v Speaker 4>for a really long time. It's standard process. It's how

0:02:08.880 --> 0:02:12.560
<v Speaker 4>state of the art improvements are diffused throughout the system,

0:02:13.000 --> 0:02:17.400
<v Speaker 4>you know, through academia and open source. But Open Ai

0:02:17.760 --> 0:02:22.280
<v Speaker 4>and Anthropic have started using this term adversarial distillation and

0:02:22.960 --> 0:02:28.360
<v Speaker 4>they are like seeking government US government protection from this process,

0:02:28.480 --> 0:02:30.520
<v Speaker 4>and they're alleging, you know that a lot of the

0:02:30.560 --> 0:02:34.560
<v Speaker 4>gains that we're seeing in Chinese models in open source

0:02:34.560 --> 0:02:37.919
<v Speaker 4>AI in particular is coming at the you know, at

0:02:37.919 --> 0:02:40.160
<v Speaker 4>the expense of anthropic.

0:02:39.960 --> 0:02:43.079
<v Speaker 1>Yeah, I have a serial distillation. It sounds like makes

0:02:43.080 --> 0:02:44.720
<v Speaker 1>me think of prohibition actually.

0:02:44.440 --> 0:02:44.520
<v Speaker 4>But.

0:02:46.440 --> 0:02:47.040
<v Speaker 2>Yeah, I was.

0:02:47.120 --> 0:02:50.480
<v Speaker 4>I was at this National Security AI conference and the

0:02:50.560 --> 0:02:56.440
<v Speaker 4>person from open ai brought it up, and Irene's salimon

0:02:56.480 --> 0:02:58.760
<v Speaker 4>from hugging face like Brazer Hins. She was like, I've

0:02:58.800 --> 0:03:01.320
<v Speaker 4>never I've been in this indust for so long. She

0:03:01.360 --> 0:03:03.360
<v Speaker 4>was a super early open ai employee, and she was like,

0:03:03.400 --> 0:03:06.200
<v Speaker 4>I've never heard this term before, but now it's now.

0:03:06.200 --> 0:03:09.799
<v Speaker 4>It's kind of it is diffusing throughout the US because,

0:03:10.400 --> 0:03:13.840
<v Speaker 4>you know, they're extremely concerned that the Chinese open source

0:03:13.919 --> 0:03:18.840
<v Speaker 4>models are cheaper, more accessible, and are being adopted very

0:03:18.840 --> 0:03:22.640
<v Speaker 4>widely by US companies. Not that they the US companies

0:03:22.680 --> 0:03:25.840
<v Speaker 4>necessarily want to talk about it, but you know, particularly

0:03:26.200 --> 0:03:31.799
<v Speaker 4>around the restrictions around fable that happened with US government,

0:03:32.040 --> 0:03:35.720
<v Speaker 4>the inability to access mythos. You know, companies are like,

0:03:35.760 --> 0:03:38.600
<v Speaker 4>it's more reliable and more cost effective for me to

0:03:38.640 --> 0:03:42.680
<v Speaker 4>build on top of Chinese technology. And I think another

0:03:43.000 --> 0:03:46.680
<v Speaker 4>kind of distinction that's really been alided over is that

0:03:47.200 --> 0:03:51.000
<v Speaker 4>distillation might be a violation of your terms of service.

0:03:51.280 --> 0:03:54.280
<v Speaker 4>You know, if you use if you access the Anthropic API,

0:03:54.480 --> 0:03:58.680
<v Speaker 4>you are not supposed to try to ask questions twenty

0:03:58.720 --> 0:04:01.080
<v Speaker 4>seven million times and then and you know, use it

0:04:01.120 --> 0:04:02.280
<v Speaker 4>to improve your own model.

0:04:02.440 --> 0:04:05.760
<v Speaker 1>What's the difference between what's happening right now in your

0:04:05.840 --> 0:04:09.840
<v Speaker 1>story and the deep Seek freak of early last year, Like,

0:04:09.920 --> 0:04:11.840
<v Speaker 1>is this kind of a continuation of the same story

0:04:11.920 --> 0:04:13.160
<v Speaker 1>or is it something slightly different?

0:04:13.560 --> 0:04:16.000
<v Speaker 4>No, I would say it's a continuation of the same story.

0:04:16.000 --> 0:04:19.560
<v Speaker 4>In that case, it was open Ai alleging that, you

0:04:19.600 --> 0:04:22.599
<v Speaker 4>know that Deep Deep Seek did the distillation, but it

0:04:22.760 --> 0:04:24.919
<v Speaker 4>also has to do with the fact that the US

0:04:25.040 --> 0:04:28.159
<v Speaker 4>was really caught off guard by how good the Deep

0:04:28.200 --> 0:04:31.960
<v Speaker 4>Seek model was. And you know, many of the gains

0:04:32.000 --> 0:04:36.560
<v Speaker 4>that you're seeing in the Chinese market are you know,

0:04:36.640 --> 0:04:39.720
<v Speaker 4>like open Ai and Anthropic would like you to believe

0:04:39.720 --> 0:04:43.200
<v Speaker 4>that it's only because they're taking from US models, but

0:04:43.240 --> 0:04:46.440
<v Speaker 4>that's not true. In fact, like many of the you know,

0:04:46.440 --> 0:04:48.719
<v Speaker 4>speaking to two experts in the field, they say that

0:04:48.839 --> 0:04:51.800
<v Speaker 4>like many of the gains that the Chinese models have

0:04:51.880 --> 0:04:57.680
<v Speaker 4>made in efficiency, in size, cost effectiveness, those have all

0:04:57.720 --> 0:05:01.120
<v Speaker 4>come from the constraints on the industry, like the fact

0:05:01.120 --> 0:05:04.680
<v Speaker 4>that we have export controls on chips. You know, they

0:05:04.680 --> 0:05:06.760
<v Speaker 4>don't have access to as many as.

0:05:06.640 --> 0:05:09.520
<v Speaker 1>Resources Jnsui Aguen, right, which is a necessity is the

0:05:09.560 --> 0:05:11.440
<v Speaker 1>mother invention and actually restricts exactly.

0:05:11.480 --> 0:05:13.960
<v Speaker 4>I mean, you know this is like the startup way.

0:05:14.080 --> 0:05:16.320
<v Speaker 4>And obviously you know that if you look at like

0:05:16.360 --> 0:05:19.200
<v Speaker 4>how much money has been raised by the major AI labs,

0:05:19.240 --> 0:05:23.560
<v Speaker 4>that's not been a constraint for them, right, And uh,

0:05:23.960 --> 0:05:28.200
<v Speaker 4>you know what's happened recently is that a one of

0:05:28.240 --> 0:05:32.760
<v Speaker 4>the Anthropic users in Hong Kong discovered that Anthropic had

0:05:33.279 --> 0:05:37.359
<v Speaker 4>was using this kind of covert way to identify Chinese users.

0:05:37.760 --> 0:05:40.080
<v Speaker 4>They were looking at what time zone they're in. They

0:05:40.080 --> 0:05:43.080
<v Speaker 4>were looking at the you know, where they're logging in from.

0:05:43.200 --> 0:05:46.680
<v Speaker 4>Is it associated? Is that domain associated with the Chinese

0:05:46.760 --> 0:05:49.800
<v Speaker 4>AI lab? And then it's sending the information back to

0:05:49.880 --> 0:05:54.760
<v Speaker 4>Anthropic but in a way it like hidden in today's date.

0:05:55.520 --> 0:05:58.920
<v Speaker 4>So like the apostrophe in today's date said like you know,

0:06:00.120 --> 0:06:04.640
<v Speaker 4>Chinese user associated with minimax logging in in this time zone,

0:06:05.200 --> 0:06:09.840
<v Speaker 4>et cetera. And people were really upset about this, like.

0:06:09.800 --> 0:06:13.039
<v Speaker 1>This is sort of a nightmare scenario for AI companies,

0:06:13.040 --> 0:06:16.080
<v Speaker 1>for users to feel like they're being actively surveiled, right,

0:06:16.120 --> 0:06:18.359
<v Speaker 1>I mean, Taylor, what's your I know that we're going

0:06:18.400 --> 0:06:20.480
<v Speaker 1>to talk about surveillance in the second half of the episode,

0:06:20.480 --> 0:06:23.080
<v Speaker 1>but what's your what's your take on the idea that

0:06:23.200 --> 0:06:26.440
<v Speaker 1>you know, Anthropic have been caught essentially spying on users

0:06:26.440 --> 0:06:28.160
<v Speaker 1>to try and figure out where they are and if

0:06:28.200 --> 0:06:31.120
<v Speaker 1>they're Chinese users, illicitly trying to distill the models.

0:06:31.360 --> 0:06:34.880
<v Speaker 3>I think this whole thing is honestly, I don't want

0:06:34.920 --> 0:06:36.680
<v Speaker 3>to say a losing game, but it seems ridiculous and

0:06:36.680 --> 0:06:39.719
<v Speaker 3>I'm kind of curious to Tasha, like why why isn't

0:06:39.760 --> 0:06:43.599
<v Speaker 3>the US developing open source models? Like why why are

0:06:43.800 --> 0:06:46.520
<v Speaker 3>you know, why can't we have our own open source

0:06:46.520 --> 0:06:48.760
<v Speaker 3>that competes with these Chinese open source models.

0:06:49.680 --> 0:06:52.960
<v Speaker 4>Well, now there is a lot more investment going into it,

0:06:53.000 --> 0:06:58.279
<v Speaker 4>but yeah, it wasn't like fine as financially lucrative. And

0:06:59.279 --> 0:07:01.719
<v Speaker 4>you know, even the you hear a lot of rhetoric

0:07:01.800 --> 0:07:04.880
<v Speaker 4>from the VCS, right like we we can't stop AI,

0:07:05.120 --> 0:07:07.559
<v Speaker 4>like we have to support open source. You weren't seeing

0:07:07.640 --> 0:07:11.360
<v Speaker 4>commensurate investment in open source. There are like a lot

0:07:11.360 --> 0:07:14.960
<v Speaker 4>of Alan Institute for AI hugging face.

0:07:16.240 --> 0:07:18.600
<v Speaker 2>There was this project called the Adam Project.

0:07:19.240 --> 0:07:23.280
<v Speaker 4>People were really interested in pushing this and having you know,

0:07:23.320 --> 0:07:25.280
<v Speaker 4>the US be the center of open source. But yeah,

0:07:25.280 --> 0:07:28.240
<v Speaker 4>there wasn't as much investment in it, and the vcs

0:07:28.280 --> 0:07:33.920
<v Speaker 4>are all backing the closed sourced companies, and so yeah,

0:07:33.960 --> 0:07:36.040
<v Speaker 4>this is the situation we now find ourselves in where

0:07:36.040 --> 0:07:37.040
<v Speaker 4>we're playing ketchup.

0:07:37.480 --> 0:07:39.960
<v Speaker 3>Yeah, I feel like we're playing ketchup. And also it's

0:07:40.000 --> 0:07:42.560
<v Speaker 3>a losing battle. Like I mean, I know Alex Karp

0:07:42.720 --> 0:07:46.080
<v Speaker 3>was talking about this recently, and obviously he's self interested.

0:07:46.160 --> 0:07:51.360
<v Speaker 3>I think volunteer uses, maybe the open source models, but

0:07:51.360 --> 0:07:55.040
<v Speaker 3>but like it just seems like this idea of this

0:07:55.200 --> 0:07:58.320
<v Speaker 3>like closed source situation and like tokens and I don't

0:07:58.360 --> 0:08:01.320
<v Speaker 3>I don't know, like everything for profit like that Anthropic

0:08:01.400 --> 0:08:04.160
<v Speaker 3>and OPA are built on is just again it's just

0:08:04.200 --> 0:08:08.000
<v Speaker 3>a losing game because ultimately open source is going to win.

0:08:08.840 --> 0:08:11.360
<v Speaker 1>I guess if Reid was here, if our play read,

0:08:11.400 --> 0:08:13.280
<v Speaker 1>because he's not here this week he's in Boston. But

0:08:13.320 --> 0:08:15.640
<v Speaker 1>I mean, I think what he would say is yes,

0:08:15.960 --> 0:08:20.200
<v Speaker 1>except that if you are leading edge company or startup

0:08:20.280 --> 0:08:22.000
<v Speaker 1>or lab, you need to have access to the best

0:08:22.040 --> 0:08:25.720
<v Speaker 1>models otherwise you will fall behind competitively. So like the

0:08:25.760 --> 0:08:28.080
<v Speaker 1>future general you say, I may be more and more

0:08:28.120 --> 0:08:30.800
<v Speaker 1>open source, but there will always be demand for like

0:08:31.240 --> 0:08:33.120
<v Speaker 1>mythos or whatever it is that you have to pay for.

0:08:33.520 --> 0:08:37.600
<v Speaker 1>Is that I mean? I guess what Red would say, Well.

0:08:37.440 --> 0:08:39.560
<v Speaker 4>You can, I mean, you could build a state of

0:08:39.559 --> 0:08:42.360
<v Speaker 4>the art model on top of open source. You don't

0:08:42.400 --> 0:08:46.880
<v Speaker 4>need to start with Anthropic. You know, Anthropic could be

0:08:46.880 --> 0:08:51.800
<v Speaker 4>building on top of deep seek. I think, like, you know,

0:08:51.840 --> 0:08:54.760
<v Speaker 4>what's really interesting is that the US position, at least

0:08:54.840 --> 0:08:57.040
<v Speaker 4>under David Sachs, was that you want to build on

0:08:57.080 --> 0:09:00.160
<v Speaker 4>top of the American stack, like you want to make

0:09:00.240 --> 0:09:03.120
<v Speaker 4>sure just in the same way that you know, China

0:09:03.120 --> 0:09:07.000
<v Speaker 4>has diffused its power by working with you know, companies

0:09:07.000 --> 0:09:09.120
<v Speaker 4>in the Middle East and Africa. They were saying, you know,

0:09:09.200 --> 0:09:12.800
<v Speaker 4>we want the US, the US stack to be like

0:09:12.920 --> 0:09:16.520
<v Speaker 4>the layer on open source. But then the Trump administration

0:09:16.600 --> 0:09:20.360
<v Speaker 4>policies are really going against that. And actually Anthropic had said,

0:09:20.720 --> 0:09:22.920
<v Speaker 4>you know, they tried to argue that if the US

0:09:23.000 --> 0:09:27.880
<v Speaker 4>enforces you know, a stronger stance against distillation, that maybe

0:09:27.920 --> 0:09:30.840
<v Speaker 4>the US labs could maintain their lead for twelve to

0:09:30.880 --> 0:09:36.760
<v Speaker 4>twenty four months longer. And yeah, it's just it's uh

0:09:37.320 --> 0:09:40.040
<v Speaker 4>has led to exactly like what you're you, Taylor. You

0:09:40.080 --> 0:09:44.800
<v Speaker 4>mentioned Alex Carp's rant recently, and it's completely related because

0:09:45.120 --> 0:09:48.800
<v Speaker 4>companies are extremely frustrated. They feel like, you know, do

0:09:48.840 --> 0:09:52.720
<v Speaker 4>they even own the intellectual property that comes from using

0:09:52.760 --> 0:09:55.600
<v Speaker 4>your model? Why are they paying for tokens? Like it's

0:09:55.640 --> 0:09:59.680
<v Speaker 4>really had people even kind of rethinking the potential business

0:09:59.760 --> 0:10:03.320
<v Speaker 4>model and what the value is for the companies if

0:10:03.400 --> 0:10:06.640
<v Speaker 4>like they don't own the technology there. Yeah, it's just

0:10:07.160 --> 0:10:09.800
<v Speaker 4>I think they feel like it's highly inefficient and not

0:10:09.880 --> 0:10:14.320
<v Speaker 4>a good way to to like start the new AI

0:10:14.440 --> 0:10:17.439
<v Speaker 4>era building on this technology that could be turned off

0:10:17.440 --> 0:10:20.160
<v Speaker 4>by the US government or anthropic could change something, or

0:10:20.200 --> 0:10:21.160
<v Speaker 4>they're spying on you.

0:10:21.520 --> 0:10:24.240
<v Speaker 3>Yeah, it seems like ridiculous. It seems like obviously we

0:10:24.280 --> 0:10:27.280
<v Speaker 3>should be moving towards an open source model. Obviously China

0:10:27.360 --> 0:10:30.199
<v Speaker 3>has the superior sort of structure to the AI environment. Also,

0:10:30.240 --> 0:10:33.200
<v Speaker 3>it's better for consumers, it's more affordable, it's better for

0:10:33.240 --> 0:10:36.199
<v Speaker 3>business consumers, better for I would say average consumers as well.

0:10:37.360 --> 0:10:39.920
<v Speaker 3>I think, I mean, I don't use deep Seek for

0:10:40.040 --> 0:10:42.640
<v Speaker 3>like tons of stuff, but I find it completely comparable

0:10:42.720 --> 0:10:46.240
<v Speaker 3>to the US you know, companies that are out there,

0:10:47.160 --> 0:10:49.480
<v Speaker 3>and I just I also wonder if if it would

0:10:49.559 --> 0:10:51.680
<v Speaker 3>help if we if we were investing more in open

0:10:51.679 --> 0:10:54.240
<v Speaker 3>source as well, if it would help sort of stem

0:10:54.280 --> 0:10:57.760
<v Speaker 3>some of the like curb some of the Antai backlash,

0:10:57.800 --> 0:10:59.800
<v Speaker 3>because I feel like so much of the Antai backlash

0:10:59.880 --> 0:11:02.760
<v Speaker 3>is not just from companies, like people like Alex Karp

0:11:02.800 --> 0:11:06.320
<v Speaker 3>going on rants, but consumers. I think consumers feel squeezed.

0:11:06.440 --> 0:11:09.320
<v Speaker 3>They feel like these companies are just doing everything to profit,

0:11:09.480 --> 0:11:12.760
<v Speaker 3>the token maxing and minning, the sort of wars. I

0:11:13.160 --> 0:11:15.800
<v Speaker 3>just feel like if we yeah had this different sort

0:11:15.800 --> 0:11:18.440
<v Speaker 3>of open source ecosystem, it would be a radically different

0:11:18.679 --> 0:11:20.840
<v Speaker 3>tech landscape and a better one in my opinion.

0:11:21.800 --> 0:11:23.760
<v Speaker 1>If you're the US government, I mean you're looking at

0:11:23.800 --> 0:11:27.040
<v Speaker 1>you're looking at the spaces and anthropic and open AI

0:11:27.120 --> 0:11:29.880
<v Speaker 1>as your kind of national champions essentially, right, And so

0:11:29.920 --> 0:11:33.199
<v Speaker 1>if you take away the ability by letting too much

0:11:33.240 --> 0:11:36.520
<v Speaker 1>distillation happen of these companies to protect their value.

0:11:36.320 --> 0:11:39.240
<v Speaker 3>Like, but let's be clear, let's be clear, open source

0:11:39.520 --> 0:11:44.720
<v Speaker 3>like models are not only developed by distilling the anthropic

0:11:44.880 --> 0:11:47.560
<v Speaker 3>and open ai like this is what they're alleging, but like,

0:11:47.880 --> 0:11:51.560
<v Speaker 3>these open source models are also progressing at a significant rate.

0:11:51.640 --> 0:11:54.080
<v Speaker 3>This is why they're so obsessed with competing with China,

0:11:54.160 --> 0:11:56.400
<v Speaker 3>Like these open source models are getting better and better,

0:11:56.440 --> 0:11:59.280
<v Speaker 3>and it's not just because they're distilling anthropic and open AI, but.

0:11:59.400 --> 0:12:02.800
<v Speaker 1>In your store or Natasha. Chinese researchers did some research

0:12:02.840 --> 0:12:05.280
<v Speaker 1>on this and it turns out that Quen, which is

0:12:05.320 --> 0:12:08.599
<v Speaker 1>Ali Barbara's AI model, this is, according to Chinese researchers,

0:12:09.040 --> 0:12:13.160
<v Speaker 1>misidentified itself as clawed in nearly one third of all cases.

0:12:13.559 --> 0:12:14.520
<v Speaker 1>I mean, that's pretty stunning.

0:12:15.240 --> 0:12:16.600
<v Speaker 2>Yeah, that's true.

0:12:16.640 --> 0:12:20.120
<v Speaker 4>But I will just say that, you know, I have

0:12:20.200 --> 0:12:24.080
<v Speaker 4>spoken to researchers who say when they test Quad and

0:12:24.120 --> 0:12:26.680
<v Speaker 4>other US models in Chinese and they do the same

0:12:26.720 --> 0:12:29.760
<v Speaker 4>thing and they say, who are you? It's as deep seek.

0:12:30.200 --> 0:12:33.120
<v Speaker 4>So I mean, like, this is what I'm trying to

0:12:33.160 --> 0:12:36.400
<v Speaker 4>say about. You know, it's really challenging when you use

0:12:36.520 --> 0:12:40.720
<v Speaker 4>this terminology that means just like a process, a methodology,

0:12:40.760 --> 0:12:46.440
<v Speaker 4>a technique, and imply that it's necessarily, you know, a

0:12:46.520 --> 0:12:49.080
<v Speaker 4>violation in some way. It might be breaking in terms

0:12:49.080 --> 0:12:51.520
<v Speaker 4>of service, which users do all the time and which

0:12:51.600 --> 0:12:54.080
<v Speaker 4>you know gets you kicked off of the platform, but

0:12:54.320 --> 0:12:57.160
<v Speaker 4>in some cases, like there is something different going on.

0:12:57.240 --> 0:13:00.160
<v Speaker 4>It's more of like jail breaking in a way. But

0:13:00.280 --> 0:13:02.640
<v Speaker 4>when you kind of use this term so broadly, it's

0:13:02.679 --> 0:13:06.720
<v Speaker 4>really hard to tell what exactly is happening and whether

0:13:07.760 --> 0:13:10.880
<v Speaker 4>some like whether it's really an egregious violation or just

0:13:10.920 --> 0:13:12.960
<v Speaker 4>something that the companies don't really want.

0:13:13.320 --> 0:13:15.920
<v Speaker 1>Tell me about this phrase secret war in the in

0:13:15.920 --> 0:13:17.880
<v Speaker 1>the headline of your story, like how is the secret

0:13:17.920 --> 0:13:21.880
<v Speaker 1>war playing out? Like who are the combatants? And what

0:13:21.880 --> 0:13:22.800
<v Speaker 1>are they doing to each other?

0:13:23.200 --> 0:13:25.960
<v Speaker 4>Well, I would say anthropic is one of the primary ones.

0:13:26.000 --> 0:13:28.200
<v Speaker 4>I mean when Taylor is asking about like why isn't

0:13:28.200 --> 0:13:31.560
<v Speaker 4>there a lot of open source. Anthropic was a big

0:13:32.360 --> 0:13:36.240
<v Speaker 4>voice in DC starting in the Biden administration, talking about

0:13:36.280 --> 0:13:39.240
<v Speaker 4>how dangerous open source was because you wouldn't be able

0:13:39.320 --> 0:13:43.440
<v Speaker 4>to check the downstream uses as closely, like if somebody

0:13:43.520 --> 0:13:46.320
<v Speaker 4>is using it to create a bioweapon or what have you,

0:13:46.400 --> 0:13:50.400
<v Speaker 4>because you can just download open source and then kind of,

0:13:50.440 --> 0:13:53.200
<v Speaker 4>you know, fork it, use it for yourself and the

0:13:53.240 --> 0:13:56.400
<v Speaker 4>company the developers of the open source technology won't necessarily

0:13:56.480 --> 0:13:59.400
<v Speaker 4>know what's happened. But I think we've also seen that

0:13:59.640 --> 0:14:02.480
<v Speaker 4>we can don't really rely on at least, you know,

0:14:02.679 --> 0:14:07.400
<v Speaker 4>when it came to how young people were using this technology,

0:14:08.160 --> 0:14:10.200
<v Speaker 4>you know, in terms of like therapy or some of

0:14:10.200 --> 0:14:14.000
<v Speaker 4>the things that we've seen with with the mental health issues,

0:14:14.080 --> 0:14:17.120
<v Speaker 4>we haven't seen like a robust reports from the companies

0:14:17.160 --> 0:14:20.840
<v Speaker 4>about downstream usage or even when it comes to you know,

0:14:20.920 --> 0:14:23.160
<v Speaker 4>how is how is it affecting jobs? It's not like

0:14:23.800 --> 0:14:27.320
<v Speaker 4>it's transparent to the public or to regulators or to

0:14:27.440 --> 0:14:31.360
<v Speaker 4>advocates how these models are being used downstream. I mean,

0:14:31.400 --> 0:14:35.280
<v Speaker 4>they put out these reports, but it's like quite filtered.

0:14:35.400 --> 0:14:37.960
<v Speaker 4>I would say, it's really hard when you dig into

0:14:38.000 --> 0:14:40.720
<v Speaker 4>methodology to really understand how people are using it. So

0:14:40.760 --> 0:14:43.840
<v Speaker 4>only the companies have access to that information. But anthropics

0:14:43.920 --> 0:14:47.960
<v Speaker 4>argument was very persuasive because you know, they're really worried

0:14:47.960 --> 0:14:52.720
<v Speaker 4>about the use of these models for uh, you know,

0:14:52.840 --> 0:14:58.040
<v Speaker 4>cyber warfare in some way, or or bioweapons or you know,

0:14:58.400 --> 0:15:02.880
<v Speaker 4>developing novel chemicals in some way, and those like once use.

0:15:03.120 --> 0:15:05.360
<v Speaker 1>This is what secure Miss Hassabis talked about this week

0:15:05.440 --> 0:15:08.320
<v Speaker 1>right the CEO of DeepMind at Google, when he called

0:15:08.320 --> 0:15:12.280
<v Speaker 1>for a US led AI standards group to regulate frontier models,

0:15:12.880 --> 0:15:14.760
<v Speaker 1>which got quite a lot of buzz. He also talked

0:15:14.760 --> 0:15:18.480
<v Speaker 1>about quote fostering international collaborational key safety issues. I assume

0:15:18.480 --> 0:15:19.440
<v Speaker 1>he was talking about China.

0:15:20.160 --> 0:15:23.400
<v Speaker 4>Yes, that's that's the way I read it as well. Yeah,

0:15:23.440 --> 0:15:28.120
<v Speaker 4>I mean it's kind of becoming this like a uniform

0:15:28.240 --> 0:15:31.640
<v Speaker 4>voice now from because we saw Jack Clark from Aanthropics

0:15:31.680 --> 0:15:35.200
<v Speaker 4>say something similar, Dario open Ai. Think what they're pushing

0:15:35.240 --> 0:15:39.200
<v Speaker 4>for is like potentially a market basedally, like you'll have

0:15:39.280 --> 0:15:42.960
<v Speaker 4>a third party person testing. It won't be through the government,

0:15:44.240 --> 0:15:48.040
<v Speaker 4>it would be a startup or a company that has

0:15:48.080 --> 0:15:49.040
<v Speaker 4>some independence.

0:15:49.080 --> 0:15:50.720
<v Speaker 2>And yeah, I could.

0:15:50.480 --> 0:15:53.480
<v Speaker 4>See more movement on this because the situation with Mythos

0:15:53.520 --> 0:15:58.360
<v Speaker 4>and fable has been so tumultuous and not good for business.

0:15:58.600 --> 0:15:59.880
<v Speaker 1>And do you think that's a good thing, I mean

0:16:00.200 --> 0:16:02.840
<v Speaker 1>for the average person in the world if this like

0:16:03.080 --> 0:16:05.520
<v Speaker 1>safety initiative happens, or do you think the fact that

0:16:05.560 --> 0:16:09.600
<v Speaker 1>it's led by the industry itself automatically creates kind of

0:16:09.640 --> 0:16:10.600
<v Speaker 1>perverse incentives.

0:16:11.480 --> 0:16:14.000
<v Speaker 4>I mean, I would say all tech policy in the

0:16:14.120 --> 0:16:17.800
<v Speaker 4>US is led by industry. They have you know, they

0:16:18.840 --> 0:16:21.720
<v Speaker 4>they have donated to all the nonprofits, they have the

0:16:21.720 --> 0:16:27.800
<v Speaker 4>biggest teams able to brief members of Congress. You know,

0:16:28.120 --> 0:16:31.120
<v Speaker 4>we just kind of naturally default to the tech executives

0:16:31.120 --> 0:16:35.080
<v Speaker 4>to tell us how to develop the technology. And you know,

0:16:35.200 --> 0:16:38.440
<v Speaker 4>lots of areas of government that need this kind of

0:16:38.440 --> 0:16:41.840
<v Speaker 4>technical expertise have been gutted. I would say there's like

0:16:41.880 --> 0:16:44.840
<v Speaker 4>suspicion obviously towards the tech companies, but I would say

0:16:44.920 --> 0:16:47.040
<v Speaker 4>all I mean, in the fifteen years that I've been

0:16:47.080 --> 0:16:50.840
<v Speaker 4>covering tech, it's all been guided by the companies themselves,

0:16:51.520 --> 0:16:53.960
<v Speaker 4>like from the beginning stage to kind of see the ideas,

0:16:53.960 --> 0:16:57.120
<v Speaker 4>and then once the law is written, you know, have

0:16:57.200 --> 0:16:59.480
<v Speaker 4>their lobbyist tweak things line to line.

0:16:59.600 --> 0:17:04.120
<v Speaker 3>I think so terrifying because like Anthropic kind of reminds

0:17:04.160 --> 0:17:07.600
<v Speaker 3>me of Meta in this way where they will they're

0:17:07.640 --> 0:17:10.880
<v Speaker 3>obsessed with safety. They want to frame everything as doing

0:17:10.920 --> 0:17:14.840
<v Speaker 3>for safety. It's for safety for safety, and often what

0:17:14.880 --> 0:17:18.760
<v Speaker 3>those regulations ultimately do is cement their power and consolidate

0:17:18.840 --> 0:17:23.359
<v Speaker 3>power in the most powerful tech companies. I mean, I

0:17:23.359 --> 0:17:27.600
<v Speaker 3>don't know. The way that Entropic and sort of the

0:17:27.600 --> 0:17:30.480
<v Speaker 3>co founders have spoken about freedom of speech and access

0:17:30.520 --> 0:17:33.399
<v Speaker 3>to information is like deeply concerning. These are people that

0:17:35.080 --> 0:17:38.040
<v Speaker 3>you know, I just I don't trust them at all

0:17:38.160 --> 0:17:41.320
<v Speaker 3>to regulate these tech policies in ways that preserve our rights.

0:17:41.760 --> 0:17:44.800
<v Speaker 3>And I think that because they go out and they

0:17:44.840 --> 0:17:47.760
<v Speaker 3>do these pr tours and they say things that people

0:17:47.760 --> 0:17:50.800
<v Speaker 3>want to hear like oh, safety safety, child safety, whatever, whatever.

0:17:51.200 --> 0:17:53.960
<v Speaker 3>People somehow think that they're responsible, and I think that

0:17:53.960 --> 0:17:56.320
<v Speaker 3>they're not responsible. I think there's I don't know. To me,

0:17:56.359 --> 0:17:58.520
<v Speaker 3>it very much feels like this METAP playbook of like

0:17:59.680 --> 0:18:01.679
<v Speaker 3>going out and just saying a bunch of stuff and

0:18:01.880 --> 0:18:05.040
<v Speaker 3>meanwhile you're just using the rules to tighten control over

0:18:05.040 --> 0:18:08.840
<v Speaker 3>our information ecosystem, titan control over the market. It's just

0:18:08.920 --> 0:18:10.120
<v Speaker 3>like regulatory capture.

0:18:11.000 --> 0:18:13.600
<v Speaker 4>I will say, it is a step above like what

0:18:14.320 --> 0:18:18.359
<v Speaker 4>was you know, pitch during the Biden EO and what's

0:18:18.359 --> 0:18:21.199
<v Speaker 4>happening now, which is like voluntary disclosure. So it's the

0:18:21.240 --> 0:18:25.000
<v Speaker 4>companies themselves doing these tests and like if you look

0:18:25.119 --> 0:18:27.040
<v Speaker 4>very closely at the you know, every time a new

0:18:27.119 --> 0:18:29.960
<v Speaker 4>model comes out, they use a different benchmark, they use

0:18:30.000 --> 0:18:35.240
<v Speaker 4>different standards. So even just having you know, I don't

0:18:35.240 --> 0:18:37.080
<v Speaker 4>want to like stand behind it, because what if we

0:18:37.080 --> 0:18:40.080
<v Speaker 4>find out, you know, the companies are like it just

0:18:40.480 --> 0:18:43.399
<v Speaker 4>the financial incentives are so skewed towards the companies that

0:18:43.440 --> 0:18:45.720
<v Speaker 4>it would be really hard for a third party. However,

0:18:46.280 --> 0:18:48.640
<v Speaker 4>I mean even just having a standard across the board

0:18:48.680 --> 0:18:53.040
<v Speaker 4>would make it a lot easier to compare models, you know,

0:18:53.160 --> 0:18:57.199
<v Speaker 4>to like for outsiders for the public to push for

0:18:57.280 --> 0:18:59.520
<v Speaker 4>different kinds of standards, like right now, we have no

0:18:59.640 --> 0:19:04.119
<v Speaker 4>say in how they're testing, what they're testing for, you know,

0:19:04.200 --> 0:19:07.080
<v Speaker 4>and and like the kinds of language.

0:19:06.680 --> 0:19:07.919
<v Speaker 2>That they use, the methods they use.

0:19:08.000 --> 0:19:10.800
<v Speaker 4>So at least if it were a third party, you know,

0:19:10.880 --> 0:19:13.359
<v Speaker 4>we would have that approach. But that, you know, doesn't

0:19:13.520 --> 0:19:15.719
<v Speaker 4>mean that it's better if it's a market based approach

0:19:15.800 --> 0:19:19.040
<v Speaker 4>versus Some of the testing that we heard was done

0:19:19.040 --> 0:19:26.920
<v Speaker 4>through Casey, the NIST organization, which is very very very

0:19:26.920 --> 0:19:29.880
<v Speaker 4>small budget but was apparently did like some great testing

0:19:29.920 --> 0:19:35.880
<v Speaker 4>of mythos maybe and the UK the UK Ai Safety Institute,

0:19:35.880 --> 0:19:37.560
<v Speaker 4>which was one of the first to like jail break

0:19:37.600 --> 0:19:41.960
<v Speaker 4>a lot of US models. I think, like a third party, yes,

0:19:42.240 --> 0:19:44.719
<v Speaker 4>and then I guess the rest kind of depends on

0:19:44.760 --> 0:19:45.800
<v Speaker 4>the on the fine print.

0:19:46.240 --> 0:19:48.639
<v Speaker 1>Just on the topic of AI and China, Taylor, I

0:19:48.680 --> 0:19:50.560
<v Speaker 1>saw another story this week they made me think of you,

0:19:50.640 --> 0:19:55.800
<v Speaker 1>which is that China has moved to ban Chinese AI

0:19:55.920 --> 0:20:00.720
<v Speaker 1>companies developing chatbots that people establish comp onionship with or

0:20:00.720 --> 0:20:03.840
<v Speaker 1>any kind of emotional connection whatsoever. I'm curious if you

0:20:03.880 --> 0:20:05.639
<v Speaker 1>saw the story and what it made you think.

0:20:06.200 --> 0:20:10.040
<v Speaker 3>Yes, they're specifically sort of trying to forbid chatbots that

0:20:10.080 --> 0:20:14.359
<v Speaker 3>are designed for companionship. So, as you mentioned, it's like

0:20:15.200 --> 0:20:17.719
<v Speaker 3>the chatbots that are like kind of like the character

0:20:17.760 --> 0:20:22.159
<v Speaker 3>AI type ones here where they're specifically designed to provide

0:20:22.440 --> 0:20:26.399
<v Speaker 3>a companion like experience, and so they're sort of built

0:20:26.560 --> 0:20:31.320
<v Speaker 3>to you know, give more maybe emotionally resident answers. They're

0:20:31.359 --> 0:20:34.479
<v Speaker 3>you know, designed to kind of help you navigate your

0:20:34.520 --> 0:20:39.320
<v Speaker 3>life problems, somebody to talk to. You know. This is

0:20:39.440 --> 0:20:42.680
<v Speaker 3>this is obviously like a very sort of controversial topic.

0:20:43.160 --> 0:20:45.959
<v Speaker 3>I think what's so concerning is is that you know,

0:20:46.119 --> 0:20:50.480
<v Speaker 3>China has extremely authoritarian uh sort of control over the

0:20:50.520 --> 0:20:55.040
<v Speaker 3>content that comes out of the tech ecosystem there, Like

0:20:55.080 --> 0:20:57.520
<v Speaker 3>they have a lot of government oversight. I don't think

0:20:57.520 --> 0:21:00.720
<v Speaker 3>that we have any evidence like this is not a

0:21:00.800 --> 0:21:03.760
<v Speaker 3>very evidence based approach, Like they're saying that they want

0:21:03.760 --> 0:21:06.840
<v Speaker 3>to ban it so that you know, basically so that

0:21:06.880 --> 0:21:09.520
<v Speaker 3>people have more babies. And I just I don't think

0:21:09.560 --> 0:21:11.959
<v Speaker 3>that the reason people are not having more babies in

0:21:12.040 --> 0:21:15.200
<v Speaker 3>China is related to the AI chatlots at all. I

0:21:15.200 --> 0:21:17.640
<v Speaker 3>think it's just again this is as you normally hear

0:21:17.680 --> 0:21:21.000
<v Speaker 3>with them sort of these types of tech laws. It's

0:21:21.080 --> 0:21:23.960
<v Speaker 3>not really based in science, it's based more in vibes.

0:21:24.680 --> 0:21:26.399
<v Speaker 1>It is interesting those and I mean it's it's like,

0:21:26.440 --> 0:21:28.880
<v Speaker 1>you know, we have, you know, in the US all

0:21:28.920 --> 0:21:31.680
<v Speaker 1>this like handering about technology out of control and their

0:21:31.680 --> 0:21:34.159
<v Speaker 1>regulation and et cetera, et cetera, and then you know,

0:21:34.240 --> 0:21:37.280
<v Speaker 1>you have this policy in China, and it's so hard

0:21:37.320 --> 0:21:40.680
<v Speaker 1>to you know, between the between the between those two

0:21:40.760 --> 0:21:45.240
<v Speaker 1>paths figure out like how how should societies you know, Well, I.

0:21:45.160 --> 0:21:47.919
<v Speaker 3>Think what's really scary and concerning is that we're seeing

0:21:47.960 --> 0:21:51.960
<v Speaker 3>America begin to regulate our information ecosystem increasingly like China.

0:21:52.119 --> 0:21:55.160
<v Speaker 3>So we continue to pass laws and actually you'll see

0:21:55.160 --> 0:22:00.000
<v Speaker 3>people that are ostensibly liberals praising China and extremely other

0:22:00.000 --> 0:22:03.199
<v Speaker 3>authoritarian regimes. And we have like Jonathan Heights, sort of

0:22:03.200 --> 0:22:06.560
<v Speaker 3>pseudoscience guy who's amassed a big audience here as a pundit,

0:22:07.119 --> 0:22:09.840
<v Speaker 3>praising countries like Malaysia and Indonesia. You know, these are

0:22:09.880 --> 0:22:12.920
<v Speaker 3>countries that don't have for free press, that arrest people

0:22:12.920 --> 0:22:15.800
<v Speaker 3>for their online speech, and often they're working with them.

0:22:15.800 --> 0:22:20.159
<v Speaker 3>I mean, we had even lawmakers praising the UAE and

0:22:20.200 --> 0:22:23.880
<v Speaker 3>their restrictions on sort of speech online and access to information.

0:22:24.000 --> 0:22:27.960
<v Speaker 3>So I think this is really worrying. There's also a

0:22:27.960 --> 0:22:30.359
<v Speaker 3>bunch of research that has come out around sort of

0:22:30.359 --> 0:22:36.600
<v Speaker 3>information and how sort of different llms criticize government. Basically,

0:22:36.600 --> 0:22:40.200
<v Speaker 3>in countries with lower levels of media freedom, when you

0:22:40.400 --> 0:22:43.760
<v Speaker 3>query models in the language of that country about politics

0:22:43.880 --> 0:22:46.440
<v Speaker 3>or the government or whatever, you are much more likely

0:22:46.520 --> 0:22:49.679
<v Speaker 3>to get a pro regime authoritarian answer than if you

0:22:49.720 --> 0:22:52.960
<v Speaker 3>ask in a different language. So these models, and this

0:22:53.000 --> 0:22:55.359
<v Speaker 3>came out in like a paper last month. Met As

0:22:55.400 --> 0:22:58.359
<v Speaker 3>Oversight Board is doing research into this. But basically what

0:22:58.400 --> 0:23:03.400
<v Speaker 3>we realize is that like training, data is shaped by

0:23:03.520 --> 0:23:06.400
<v Speaker 3>the sort of political ecosystem of that country. This has

0:23:06.440 --> 0:23:09.520
<v Speaker 3>consequences for LLM output. We need to make sure that

0:23:09.600 --> 0:23:13.480
<v Speaker 3>we are preserving our free and open information ecosystem here

0:23:13.480 --> 0:23:17.000
<v Speaker 3>in the US. And I don't think we're doing any

0:23:17.040 --> 0:23:18.720
<v Speaker 3>of that. And I think when we look at laws

0:23:18.760 --> 0:23:21.800
<v Speaker 3>like what China is doing and we sort of praise it,

0:23:21.600 --> 0:23:22.920
<v Speaker 3>it makes me very nervous.

0:23:28.960 --> 0:23:31.480
<v Speaker 1>When we come back, Tayla tells us why all the

0:23:31.720 --> 0:23:47.639
<v Speaker 1>girls having a hot surveillance summer stay with us? Welcome back, Taylor.

0:23:47.920 --> 0:23:50.639
<v Speaker 1>Where does the phrase hot surveillance summer come from? What

0:23:50.720 --> 0:23:51.200
<v Speaker 1>does it mean?

0:23:52.840 --> 0:23:57.280
<v Speaker 3>So it's actually from this podcaster Natalie, who hosts this

0:23:57.320 --> 0:24:00.560
<v Speaker 3>podcast Boys Club It's like a tech podcast. She wrote

0:24:00.600 --> 0:24:03.600
<v Speaker 3>this post about having a surveillance summer that was basically

0:24:03.640 --> 0:24:09.680
<v Speaker 3>pro Meta smart glasses. She came out right along when

0:24:09.880 --> 0:24:12.320
<v Speaker 3>sort of Meta began to launch this campaign with Kylie

0:24:12.400 --> 0:24:15.040
<v Speaker 3>Jenner and all the influencers are having, you know, their

0:24:15.080 --> 0:24:18.159
<v Speaker 3>sort of surveillance summer moment. Basically what she said is

0:24:18.200 --> 0:24:20.840
<v Speaker 3>like masurveillances everywhere, so why not participate in it?

0:24:22.320 --> 0:24:23.720
<v Speaker 2>Which is I.

0:24:23.760 --> 0:24:28.800
<v Speaker 3>Trust, it's horrifying. I hate it. But a lot of influencers,

0:24:28.800 --> 0:24:31.320
<v Speaker 3>if you look at the replies to her posts, were like, yes,

0:24:31.640 --> 0:24:33.960
<v Speaker 3>let's go. And you know, I was in New York

0:24:34.440 --> 0:24:36.399
<v Speaker 3>just last week and I saw the first pair of

0:24:36.440 --> 0:24:38.520
<v Speaker 3>Meta smart glasses in the wild on this girl. I

0:24:38.560 --> 0:24:39.960
<v Speaker 3>don't know if she was an influencer, but she was

0:24:40.000 --> 0:24:42.719
<v Speaker 3>really pretty, and I imagine that she was an influencer

0:24:42.720 --> 0:24:46.360
<v Speaker 3>because she had the early access to these glasses. But they're, yeah,

0:24:46.359 --> 0:24:49.280
<v Speaker 3>they're getting popular and people just kind of are treating

0:24:49.280 --> 0:24:51.560
<v Speaker 3>it as this like fun thing and they love the

0:24:51.640 --> 0:24:56.040
<v Speaker 3>Kylie glasses and yeah, they're talking about sort of proactively

0:24:56.040 --> 0:24:58.000
<v Speaker 3>having a hot surveillanceummer in like a positive way.

0:24:58.480 --> 0:25:00.880
<v Speaker 1>Tally can you kind of lay out the history smart glasses.

0:25:00.880 --> 0:25:03.439
<v Speaker 1>I remember when I was a TV producer back in

0:25:03.480 --> 0:25:07.399
<v Speaker 1>like the twenty eleven period, Sebastian Thrun came on the

0:25:07.440 --> 0:25:11.040
<v Speaker 1>show I worked, only had this like Bionic Man prototype

0:25:11.040 --> 0:25:15.000
<v Speaker 1>of the Google glass that didn't quite catch on. Then

0:25:15.040 --> 0:25:18.560
<v Speaker 1>they were the first generation of Snapchat spectacles. Then metas

0:25:18.640 --> 0:25:21.200
<v Speaker 1>ray Band partnership kind of brought these somewhat into the mainstream.

0:25:21.240 --> 0:25:25.600
<v Speaker 1>Obviously Kylie is now has her own Meta ray Ban collab,

0:25:26.000 --> 0:25:27.640
<v Speaker 1>and then later this year or early next year, snap

0:25:27.760 --> 0:25:30.480
<v Speaker 1>had to going to release these new spectacles are apparently

0:25:30.480 --> 0:25:32.280
<v Speaker 1>going to cost two thousand dollars. I mean, what's the

0:25:32.359 --> 0:25:35.159
<v Speaker 1>kind of what's the technological history of these glasses? And

0:25:35.280 --> 0:25:37.000
<v Speaker 1>are they are there things I've actually I'm not sure

0:25:37.040 --> 0:25:40.399
<v Speaker 1>i've ever seen anyone actually wearing them in the wild

0:25:40.680 --> 0:25:41.560
<v Speaker 1>as well as I know.

0:25:41.920 --> 0:25:47.960
<v Speaker 3>I've seen them unfortunately never Google glass. But but but yeah,

0:25:48.000 --> 0:25:50.360
<v Speaker 3>I mean these, I mean Meta RayBan glasses are popular.

0:25:50.600 --> 0:25:52.320
<v Speaker 3>I mean people have them. I've seen them in you

0:25:52.760 --> 0:25:56.600
<v Speaker 3>in LA. I think what's interesting to me, I mean

0:25:56.640 --> 0:25:59.080
<v Speaker 3>you just sort of like laid out the general progression

0:25:59.200 --> 0:26:01.439
<v Speaker 3>of a smart glas is I think the idea of

0:26:01.480 --> 0:26:04.040
<v Speaker 3>having something on your face that is constantly recording. That

0:26:04.119 --> 0:26:07.480
<v Speaker 3>can you know, act as this sort of second brain

0:26:07.560 --> 0:26:09.800
<v Speaker 3>for you, remind you of someone you can't remember, someone

0:26:09.880 --> 0:26:11.320
<v Speaker 3>you know, Like that's always been the sort of like

0:26:11.359 --> 0:26:15.280
<v Speaker 3>sci fi fantasy for years. We've had recording capabilities. So

0:26:15.560 --> 0:26:19.399
<v Speaker 3>I actually had the Snapchat spectacles back in twenty sixteen,

0:26:19.920 --> 0:26:21.920
<v Speaker 3>and you know, you could record very easily, you could

0:26:21.920 --> 0:26:24.520
<v Speaker 3>post a snapchat, so and that is sort of what

0:26:24.560 --> 0:26:27.560
<v Speaker 3>pioneered I think the popularity, the niche popularity, I guess,

0:26:27.600 --> 0:26:30.800
<v Speaker 3>of the original sort of meta rap ban glasses. What

0:26:30.960 --> 0:26:32.920
<v Speaker 3>makes this moment interesting to me and why I think

0:26:32.920 --> 0:26:37.000
<v Speaker 3>we're reaching a turning point in the normalization of these

0:26:37.040 --> 0:26:40.880
<v Speaker 3>products is that you now have sort of plausible deniability.

0:26:40.920 --> 0:26:43.880
<v Speaker 3>Previously they were only for recording, right, And so it's

0:26:43.920 --> 0:26:46.920
<v Speaker 3>kind of weird, like your people know that that's why

0:26:46.960 --> 0:26:49.080
<v Speaker 3>you're wearing them. It's like, what do you creep? Why

0:26:49.119 --> 0:26:50.680
<v Speaker 3>do you have to record? Why you know? Why can't

0:26:50.680 --> 0:26:52.200
<v Speaker 3>you just take out your phone like a normal person.

0:26:52.920 --> 0:26:55.120
<v Speaker 3>But two things have happened. One, we've actually normalized taking

0:26:55.160 --> 0:26:57.560
<v Speaker 3>out our phone and recording at all times. That's basically

0:26:57.560 --> 0:27:01.159
<v Speaker 3>been completely destigmatized ten years ago. That was actually still stigmatized.

0:27:01.160 --> 0:27:02.760
<v Speaker 3>Now there was like houn its like influencers in the

0:27:02.760 --> 0:27:04.879
<v Speaker 3>wild that would like make fun of people recording in public.

0:27:05.000 --> 0:27:09.080
<v Speaker 3>Now everyone records in public. Second of all is that

0:27:09.160 --> 0:27:11.520
<v Speaker 3>these glasses have AI integrated into them, and so they're

0:27:11.560 --> 0:27:14.439
<v Speaker 3>being sold as this like smart AI assistant. And so

0:27:14.480 --> 0:27:17.200
<v Speaker 3>the people that I spoke to in the replies of

0:27:17.240 --> 0:27:19.199
<v Speaker 3>this tweet that I was like, why would you do this,

0:27:19.359 --> 0:27:21.919
<v Speaker 3>They're like, well, I want AI. I want somebody to

0:27:22.080 --> 0:27:24.359
<v Speaker 3>see that everything that I see and help me. And

0:27:24.400 --> 0:27:27.080
<v Speaker 3>that is I think when I sort of see the

0:27:27.160 --> 0:27:31.160
<v Speaker 3>discussion around these consumer products, they're like, well, yeah, people

0:27:31.160 --> 0:27:33.080
<v Speaker 3>that record all day around them? Are you know, if

0:27:33.080 --> 0:27:35.080
<v Speaker 3>you're sort of like recording women with your glasses, you're

0:27:35.080 --> 0:27:37.680
<v Speaker 3>a greep. But I actually use them for help for AI.

0:27:37.920 --> 0:27:40.400
<v Speaker 3>And this is like part of the sort of girl

0:27:40.440 --> 0:27:42.960
<v Speaker 3>bossification of AI, right, Like I use my metal smart

0:27:42.960 --> 0:27:45.400
<v Speaker 3>glasses all day long and they help me figure out

0:27:45.400 --> 0:27:48.840
<v Speaker 3>what to buy at arrow on or you know, I

0:27:48.840 --> 0:27:53.000
<v Speaker 3>don't know, remember to shop stop by Alo or something.

0:27:53.040 --> 0:27:54.960
<v Speaker 3>I don't know. But it's yeah, which is I think

0:27:54.960 --> 0:27:57.080
<v Speaker 3>notable because when you look at who they've been giving

0:27:57.119 --> 0:27:58.600
<v Speaker 3>these glasses to, it is a lot of the like

0:27:58.640 --> 0:28:02.679
<v Speaker 3>West Village girlies, like lifestyle influencers, these aspirational women.

0:28:03.359 --> 0:28:05.840
<v Speaker 1>Natasha, you've been flowing this story, yes.

0:28:05.720 --> 0:28:08.480
<v Speaker 4>I mean like I of course went through the whole

0:28:09.080 --> 0:28:11.840
<v Speaker 4>Kylie photodump of the of the Meta glasses. And I

0:28:11.840 --> 0:28:14.520
<v Speaker 4>think when I moved to San Francisco was just when,

0:28:15.400 --> 0:28:20.240
<v Speaker 4>like maybe slightly after, people were being accosted in bars

0:28:20.280 --> 0:28:23.199
<v Speaker 4>for wearing the Google glasses, which are you know, like

0:28:23.440 --> 0:28:27.480
<v Speaker 4>extremely it looked you can tell when somebody was wearing

0:28:27.520 --> 0:28:29.639
<v Speaker 4>Google glasses, So I did see them here. I have

0:28:29.800 --> 0:28:32.720
<v Speaker 4>only seen the Metal ones in La one time, and

0:28:32.760 --> 0:28:35.720
<v Speaker 4>people were giving the guy a hard time. He was

0:28:35.760 --> 0:28:38.200
<v Speaker 4>on it was like on Abbot Kinney and you know,

0:28:38.360 --> 0:28:39.520
<v Speaker 4>it was just like what are you doing?

0:28:39.640 --> 0:28:43.400
<v Speaker 2>U creep? But this was maybe in January, so a

0:28:43.440 --> 0:28:44.040
<v Speaker 2>while ago.

0:28:45.320 --> 0:28:53.280
<v Speaker 4>Yeah, I mean, I understand the appeal of potentially having

0:28:53.960 --> 0:28:59.880
<v Speaker 4>something like as an always on assistant. I for some reason,

0:28:59.920 --> 0:29:02.400
<v Speaker 4>I always think about that scene in Devilwaar's Prada where

0:29:02.400 --> 0:29:05.680
<v Speaker 4>they're like whispering the name of the person to you.

0:29:06.280 --> 0:29:09.680
<v Speaker 4>I understand why we would want to feel like we're

0:29:09.880 --> 0:29:13.240
<v Speaker 4>the boss and we have you know, minions helping us.

0:29:13.280 --> 0:29:15.480
<v Speaker 2>But I don't know. I mean at least when I

0:29:15.520 --> 0:29:15.920
<v Speaker 2>put on.

0:29:15.960 --> 0:29:18.200
<v Speaker 4>The Google glasses, I remember too that it was just

0:29:18.520 --> 0:29:22.680
<v Speaker 4>it was not the seamless minority report experience that that

0:29:22.760 --> 0:29:23.880
<v Speaker 4>Sci Fi promised us.

0:29:23.960 --> 0:29:26.000
<v Speaker 2>So I don't know.

0:29:26.080 --> 0:29:28.960
<v Speaker 4>I mean, what do you think Taylor about, like how

0:29:30.480 --> 0:29:34.600
<v Speaker 4>how savvy they've been with targeting the right influencers. Do

0:29:34.640 --> 0:29:39.479
<v Speaker 4>you feel like going through women has been like, has

0:29:39.560 --> 0:29:41.280
<v Speaker 4>led to more adoption than they would have if they

0:29:41.360 --> 0:29:43.240
<v Speaker 4>tried to go like the Google glass route.

0:29:43.640 --> 0:29:46.960
<v Speaker 3>Undeniably. I think, like number one, they've destigmatized the idea

0:29:47.000 --> 0:29:49.360
<v Speaker 3>that these are creep glasses used to surveil women, which

0:29:49.400 --> 0:29:54.280
<v Speaker 3>is what people sort of thought of metaglasses before. And

0:29:54.800 --> 0:29:57.480
<v Speaker 3>I think they've really made it aspirational. I think the

0:29:57.960 --> 0:30:01.479
<v Speaker 3>HIGHI partnership was brilliant. I think the sort of rollout

0:30:01.480 --> 0:30:03.960
<v Speaker 3>has been great, and I think you're seeing a lot

0:30:03.960 --> 0:30:07.600
<v Speaker 3>of women get reflexively defensive and be like, well I

0:30:07.640 --> 0:30:09.400
<v Speaker 3>want I mean I talked to my friend the other

0:30:09.440 --> 0:30:11.320
<v Speaker 3>day who wants to buy them because she has her kids,

0:30:11.320 --> 0:30:13.120
<v Speaker 3>and she's like, oh, it would be so great, Like

0:30:13.160 --> 0:30:14.840
<v Speaker 3>my hands are always full, I could just talk to

0:30:14.880 --> 0:30:17.280
<v Speaker 3>the glasses and also then I could record moments with

0:30:17.320 --> 0:30:18.800
<v Speaker 3>my children. I don't have to have my phone out

0:30:18.800 --> 0:30:21.640
<v Speaker 3>and like, I totally get it. I totally get it.

0:30:22.000 --> 0:30:22.400
<v Speaker 4>Like I.

0:30:24.360 --> 0:30:26.560
<v Speaker 3>Hate the mass surveillance aspect. I think like, if we

0:30:26.600 --> 0:30:28.680
<v Speaker 3>are going to normalize this level of masurveillance, let's have

0:30:28.720 --> 0:30:32.400
<v Speaker 3>like serious conversations around policy and privacy and privacy protections.

0:30:32.400 --> 0:30:34.080
<v Speaker 3>I think it's interesting as well. There was that Wired

0:30:34.120 --> 0:30:37.160
<v Speaker 3>report about that system called face Card that Meta was

0:30:37.440 --> 0:30:38.400
<v Speaker 3>going to ask you about this.

0:30:38.400 --> 0:30:41.640
<v Speaker 1>This is interesting idea that the lasses were cross reference

0:30:41.680 --> 0:30:43.760
<v Speaker 1>with the database of all faces, so you could kind

0:30:43.760 --> 0:30:47.200
<v Speaker 1>of live identify people. But then they discontinued that, or

0:30:47.240 --> 0:30:49.480
<v Speaker 1>they kind of sum set it for the time being

0:30:49.520 --> 0:30:51.520
<v Speaker 1>off the wide article or what happened.

0:30:51.440 --> 0:30:55.360
<v Speaker 3>Well sort of, And you know, Boz was on this

0:30:55.400 --> 0:30:59.840
<v Speaker 3>podcast with Nicholas Thompson, the editor of The Atlantic, a

0:31:00.160 --> 0:31:03.040
<v Speaker 3>Meta executive, and he's like, you know, I can't believe

0:31:03.040 --> 0:31:04.640
<v Speaker 3>there was so much backlash. Like, first of all, he

0:31:04.680 --> 0:31:07.800
<v Speaker 3>confirmed the story after Andy Stone, the comms person for Meta,

0:31:07.840 --> 0:31:10.280
<v Speaker 3>was like on Twitter all day denying it. This is nonsense,

0:31:10.320 --> 0:31:12.880
<v Speaker 3>this is nonsense. Then you have bas an executive going

0:31:12.920 --> 0:31:14.960
<v Speaker 3>on a podcast and being like, actually, we're really proud

0:31:14.960 --> 0:31:17.680
<v Speaker 3>of this work and it's actually really amazing. My favorite

0:31:17.720 --> 0:31:19.640
<v Speaker 3>part of that interview is that he's like, oh, people

0:31:19.640 --> 0:31:22.040
<v Speaker 3>think that there's just going to be some master database

0:31:22.080 --> 0:31:23.480
<v Speaker 3>and like it's just going to be pulling everyone for

0:31:23.560 --> 0:31:26.080
<v Speaker 3>the face databases. No, it's going to be pulling every

0:31:26.240 --> 0:31:28.800
<v Speaker 3>it's going to be pulling from a database that shows

0:31:28.800 --> 0:31:31.240
<v Speaker 3>you your connections, if you met that person once or whatever.

0:31:31.280 --> 0:31:33.920
<v Speaker 3>It's just like, Okay, that is such a meaningless distinction.

0:31:33.960 --> 0:31:36.520
<v Speaker 3>The point is that you are creating a database of

0:31:36.560 --> 0:31:40.840
<v Speaker 3>biometric data and facial recognition data and tying that to people,

0:31:40.920 --> 0:31:43.400
<v Speaker 3>and that is what we are mad about, you know.

0:31:43.760 --> 0:31:46.200
<v Speaker 1>So it's amazing. I remember when Facebook first came out,

0:31:46.200 --> 0:31:47.920
<v Speaker 1>and I was at high school. Maybe not when it

0:31:47.920 --> 0:31:49.880
<v Speaker 1>first came out, but when it first started diffuse and

0:31:50.120 --> 0:31:53.040
<v Speaker 1>I was supposed sixteen or seventeen, I remember looking at

0:31:53.080 --> 0:31:55.720
<v Speaker 1>him thinking, oh my god, like there was soon not

0:31:55.840 --> 0:31:59.080
<v Speaker 1>in the not too in futury, possible to overlay everyone voluntary,

0:31:59.240 --> 0:32:02.480
<v Speaker 1>voluntarily kind of uploading themselves to this platform and just

0:32:02.600 --> 0:32:04.800
<v Speaker 1>having no expectation of privacy and people being able to

0:32:04.840 --> 0:32:07.480
<v Speaker 1>know who you are all the time. And it's like, wh.

0:32:07.560 --> 0:32:10.920
<v Speaker 3>Boy, there, which, by the way, we normalize that, I mean,

0:32:11.040 --> 0:32:14.320
<v Speaker 3>I like, I mean, I did the same thing with Facebook,

0:32:14.360 --> 0:32:16.360
<v Speaker 3>like I would stock down everyone in my class, Like

0:32:16.400 --> 0:32:19.520
<v Speaker 3>I think I added everyone in my college dorm, you

0:32:19.560 --> 0:32:21.920
<v Speaker 3>know when I got into that. And so it's like

0:32:22.040 --> 0:32:26.200
<v Speaker 3>I think that, like people used to have a huge

0:32:26.200 --> 0:32:29.920
<v Speaker 3>amount of privacy, we have watched privacy be obliterated over

0:32:30.000 --> 0:32:32.920
<v Speaker 3>the past couple decades with the rise of social media.

0:32:33.320 --> 0:32:37.960
<v Speaker 3>Now we have AI which is super charging privacy violations.

0:32:38.440 --> 0:32:41.240
<v Speaker 3>I think that we are not having any conversations about

0:32:41.360 --> 0:32:43.720
<v Speaker 3>privacy like we should be having. We should be passing

0:32:43.760 --> 0:32:48.440
<v Speaker 3>privacy laws. Instead, we're passing the opposite. We're actually mandating.

0:32:48.920 --> 0:32:52.680
<v Speaker 3>Over twenty five states have mandated identity verification laws to

0:32:52.760 --> 0:32:56.240
<v Speaker 3>use Internet and then or to use parts of the Internet,

0:32:56.280 --> 0:32:58.160
<v Speaker 3>I should say. And then I think there's forty seven

0:32:58.200 --> 0:33:00.640
<v Speaker 3>states that are that are that have total either pass

0:33:00.680 --> 0:33:02.640
<v Speaker 3>these laws or are considering past them. And then we

0:33:02.680 --> 0:33:05.560
<v Speaker 3>have the Federal Kids Act discussions. So we're about to

0:33:05.640 --> 0:33:09.880
<v Speaker 3>mandate that Meta start harvesting biometric data, including on children

0:33:10.200 --> 0:33:12.920
<v Speaker 3>while they're building this technology. So if we want, if

0:33:12.920 --> 0:33:16.360
<v Speaker 3>we're really against surveillance, then we have to have a

0:33:16.360 --> 0:33:17.720
<v Speaker 3>bigger discussion about privacy.

0:33:18.280 --> 0:33:20.280
<v Speaker 1>Just to close, Natasha, I want to ask you about

0:33:20.360 --> 0:33:22.920
<v Speaker 1>and There's two interesting stories I saw this week that

0:33:22.960 --> 0:33:26.600
<v Speaker 1>are kind of adjacent to this. One is Apple suing

0:33:26.760 --> 0:33:32.160
<v Speaker 1>open Ai for allegedly getting interviewing Apple execs and basically

0:33:32.240 --> 0:33:36.800
<v Speaker 1>using the interviews to steal Apple trade secrets around device manufacturing.

0:33:37.280 --> 0:33:39.560
<v Speaker 1>And the other was a report in Bloomberg about open

0:33:39.600 --> 0:33:42.720
<v Speaker 1>ai bringing its first device to market, which is not

0:33:42.800 --> 0:33:44.960
<v Speaker 1>a phone and not glasses. It sounds more like an

0:33:45.000 --> 0:33:47.920
<v Speaker 1>Alexa or a Sonos. But can you give us some

0:33:48.400 --> 0:33:49.640
<v Speaker 1>contexts on those two stories.

0:33:50.320 --> 0:33:55.160
<v Speaker 4>Yeah, the Apple lawsuit was super interesting because it was

0:33:55.280 --> 0:33:59.560
<v Speaker 4>chock full of details about, you know, even having access

0:33:59.640 --> 0:34:04.040
<v Speaker 4>to people's texts and messages where they're like, lol, I

0:34:04.080 --> 0:34:07.920
<v Speaker 4>guess I still have access to this or I didn't

0:34:08.000 --> 0:34:09.960
<v Speaker 4>know you could, you know, bring a piece of an

0:34:09.960 --> 0:34:13.040
<v Speaker 4>Apple device out of the out of the company because

0:34:13.560 --> 0:34:16.359
<v Speaker 4>this long time Apple executive who moved to open ai

0:34:16.560 --> 0:34:19.839
<v Speaker 4>suggested that when you come to open ai, you bring

0:34:20.000 --> 0:34:24.000
<v Speaker 4>like an actual physical thing for a little show and

0:34:24.040 --> 0:34:27.600
<v Speaker 4>tell session. But one thing I thought was also really

0:34:27.600 --> 0:34:33.480
<v Speaker 4>interesting is Tony Fidel, the CEO of Nest. He reached

0:34:33.480 --> 0:34:39.399
<v Speaker 4>out after this newsletter writer Ben Thompson Etstra Tetri wrote

0:34:39.400 --> 0:34:41.680
<v Speaker 4>about it, and he said, this is how Apple scares

0:34:41.719 --> 0:34:44.640
<v Speaker 4>employees like this is I think what he was doing

0:34:44.719 --> 0:34:47.240
<v Speaker 4>was trying to frame it as not necessarily going after

0:34:47.440 --> 0:34:50.319
<v Speaker 4>open Ai, like you know, being sure that they'll they'll

0:34:50.360 --> 0:34:54.040
<v Speaker 4>be able to you know, get damages or prevent something

0:34:54.040 --> 0:34:57.120
<v Speaker 4>from open Ai, but rather as a message to Apple employees,

0:34:57.560 --> 0:35:00.000
<v Speaker 4>of which open ai has hired four hundreds.

0:35:00.239 --> 0:35:03.520
<v Speaker 1>But these were texts onto Apple employees on their work phones,

0:35:03.560 --> 0:35:05.799
<v Speaker 1>which therefore Apple could read. Or how did Apple start

0:35:05.840 --> 0:35:07.360
<v Speaker 1>to like see this, Corsemond.

0:35:07.080 --> 0:35:10.120
<v Speaker 4>Yeah, yeah, what like the a few of the texts

0:35:10.120 --> 0:35:13.400
<v Speaker 4>that they already had were on their Apple devices remaining

0:35:13.400 --> 0:35:15.600
<v Speaker 4>on their Apple devices, and they believe that, you know,

0:35:15.640 --> 0:35:17.719
<v Speaker 4>if they take it to discovery, they'll be able to

0:35:17.800 --> 0:35:22.480
<v Speaker 4>access even more. And they also mentioned this, you know,

0:35:22.520 --> 0:35:26.080
<v Speaker 4>they mentioned the Johnny I've partnership, the Johnny Ive company

0:35:26.120 --> 0:35:29.319
<v Speaker 4>that was that was acquired by open Ai, and they

0:35:29.360 --> 0:35:32.080
<v Speaker 4>were saying that you know, Johnny went to it was

0:35:32.120 --> 0:35:36.080
<v Speaker 4>something in particular about the metal finishing, like a proprietary

0:35:36.120 --> 0:35:39.200
<v Speaker 4>metal finishing that only Apple knows, and Johnny I've it

0:35:39.239 --> 0:35:42.000
<v Speaker 4>sounded like basically he's going to their suppliers and using

0:35:42.080 --> 0:35:45.120
<v Speaker 4>like Apple terminology, and so they had the impression this

0:35:45.239 --> 0:35:49.640
<v Speaker 4>was for an Apple product, And yeah, I guess we'll

0:35:49.680 --> 0:35:52.600
<v Speaker 4>see when this new Alexa comes out. I thought it

0:35:52.640 --> 0:35:54.680
<v Speaker 4>was supposed to be that tiny thing that we saw

0:35:55.239 --> 0:35:58.319
<v Speaker 4>like a few executives carrying around and get photographed with

0:35:58.520 --> 0:36:03.759
<v Speaker 4>like a compact almost with like in a dark metal finish,

0:36:04.080 --> 0:36:06.399
<v Speaker 4>with I think a little open AI logo. It looked

0:36:06.440 --> 0:36:09.719
<v Speaker 4>like a little clamshaw kind of thing and people assumed

0:36:10.200 --> 0:36:11.480
<v Speaker 4>that that was the device.

0:36:11.600 --> 0:36:15.720
<v Speaker 2>But yeah, it sounds much less exciting.

0:36:16.360 --> 0:36:18.960
<v Speaker 1>That's all we have time for this week. Taylor, Natasha,

0:36:19.000 --> 0:36:19.439
<v Speaker 1>thank you.

0:36:19.600 --> 0:36:20.200
<v Speaker 2>Good to see you.

0:36:20.400 --> 0:36:21.879
<v Speaker 3>Thanks for having us.

0:36:43.520 --> 0:36:46.680
<v Speaker 1>For tech stuff. I'm os Voloshin. This episode was produced

0:36:46.680 --> 0:36:49.960
<v Speaker 1>by Eliza Dennis. It was executive produced by me and

0:36:50.040 --> 0:36:54.839
<v Speaker 1>Julian Nutta for Kaleidoscope and Katrina norvelf iHeart Podcasts. Jack

0:36:54.880 --> 0:36:57.960
<v Speaker 1>Insley makes this episode and Kyle Murdoch wrote our theme song.

0:36:58.440 --> 0:37:01.240
<v Speaker 1>A special thank you to Taylor Lorenz and Natasha Tiku.

0:37:01.640 --> 0:37:03.759
<v Speaker 1>Please check out all the work they put out into

0:37:03.760 --> 0:37:06.279
<v Speaker 1>the world. We're lucky to call them friends of the Pod.