Ben Whitelaw 0:05

Mike, I'm excited to tell you what my fun thing this week is on Ctrl-Alt-Speech. I'm always struggling for ideas basically for this opening, cold opening, and you

Mike Masnick 0:14

There's so much fun. There's so much fun on the internet. How, how could you struggle?

Ben Whitelaw 0:18

not seemingly where I hang out. Um, and actually my, fun thing this week is actually from LinkedIn, which is not normally a place that you'd say fun things exist.

Mike Masnick 0:28

the anti-fun. It's where fun goes to die.

Ben Whitelaw 0:31

For sure. For sure. But I, I found, I managed to find it, the one single fun LinkedIn post, by a woman called Brandi Fleming. You won't know her. I don't know her either. She's, she's a relatively regular person. She's not into online safety per se, but she did post about the presentation that she gave her 13-year-old about online safety.

Mike Masnick 0:51

Oh, okay

Ben Whitelaw 0:52

So she explained how she sat her 13-year-old down, gave him a full kind of, 10-deck slide deck in which she outlined the rules of the road when using the internet. Clearly this, child is getting access to a phone or, is, moving class. It's the start of a new term and she does a great job in a very short space of time of like outlining the responsibilities that come with using the internet. How if you see something bad, you can tell her and presumably her partner, about what's going on. There's no need to keep things kind of secret. her kind of key slide I think is like a really fun one that says like, "Things that stay private," and it includes a list of things including your body. So, you know, if there's any, lack of clarity about what this young person can put online, I think after with that presentation, they're probably much, much clearer in their head. So yeah, I thought that was a great thing to, to share with our listeners.

Mike Masnick 1:45

And that's, that's useful. That's not just like random fun. That's, actually useful

Ben Whitelaw 1:49

Yeah. What about you? What have you been paying attention to?

Mike Masnick 1:52

So I have been debating if, if my thing this week is fun or scary or terrifying or, or, or some sort of combination of that. you may have missed this. This, became very big in certain circles, but it's this open source project which was just open sourced, uh, called God's Eye View. Have you seen this?

Ben Whitelaw 2:10

Nein

Mike Masnick 2:11

Okay. So it was created by a guy who was the product manager for Google Maps, Google Earth, has, built this whole thing. So he's basically rebuilt that, as a totally open source tool, but, like, created as if you are working for whatever intelligence agency that you want that in your dreams had access to everything around the world. but it pulls in all sorts of public data. So it looks like Google Earth, where you see an Earth and you can zoom in on any place in the world, but you can click on different types of content and layers appear, and it's all pulled from public sources. But if you wanna see any, airplane flight, like, you know, that is all public data. So as you're looking at it, you see planes flying by, and you can click on an airplane. You can see which plane it is. You can switch to the cockpit view, and suddenly you're in the cockpit of that plane and seeing effectively what the pilots see. then you can turn on traffic, and it shows, like, cars moving along the thing. You can add in... I, I don't know exactly how he did it, but he has, figured out where all of the public videos are. So, like, there's, like, all the traffic cams and everything. So you wanna see what the traffic looks like on a highway, uh, near your home? You can do that just by spinning through this map, and it, it has all these different layers, and you can-- and it's all open source, and you run it locally, and you can build in AI tools. I saw him demo it where he's added a voice AI thing. So he's just speaking to it, and he's just like, At one point he was looking at, like, the fiber optic cables and shipping routes of, different ships in different parts of the world. It's-- But it's all public data.

Ben Whitelaw 3:55

Yeah, that's not fun. That's you've, you've misread the brief, Mike

Mike Masnick 4:00

d- There's an element to it that is fun because it's like, "Oh my gosh, look at this. This is kind of cool that you can look at that on your computer." But yes, there is a part of it that is absolutely terrifying. It's the kind of thing where it's like 20 years ago if this was in a movie, you would have, like, you know, the intelligence agency, like, opening up this app on their computer, and, like, everybody in the tech world would be like, "Pfft, that's not how any of this works." And now it works on all of our computers for anyone.

Ben Whitelaw 4:29

God, this episode's only gonna go downhill from there. Hello, and welcome to Ctrl Alt Speech, the podcast where we make sense of the major debates shaping online speech, platform power, content moderation, and the future of the internet. It's September the 17th, 2026, and this week we're talking about EU teen ban announcement, some naughty data on AI in education, and for our extended Patreon episode, what it actually means to pace the frontier. Talking of pacing, Ctrl Alt Speech is the slow drumbeat of weekly safety news from Mike Masnick, founder and editor of Techdirt, and Ben Whitelaw, myself, the founder and editor of Everything in Moderation. I thought you might like that one, Mike. It's, uh, I had a, I had a few others lined up there, but

Mike Masnick 5:19

No, no, that's it. You, you, you got it right. I, I always look forward, you know, more than even the fun things, I think I look forward to how you describe the podcast each week

Ben Whitelaw 5:28

Yeah, between that and the episode titles, we do have a bit of fun on it. w- we had a few kind of nice comments about, episode titles recently,

Mike Masnick 5:36

Yeah, people seem to... I, I mean, we always like it when we hear people tell us that they enjoy the work that we put into the titles, and we do put a lot of effort into them. And I think we've been, we've been on a roll lately of really good titles

Ben Whitelaw 5:48

Yeah. we don't really talk about this a lot, but we, we have a bit of a game where we try to come up with a title ourselves. you, I, and, Lee, our producer, and we do kind of ask for AI input, and I would say 9 times out of 10 we go with the episode title that we come up with, which is not

Mike Masnick 6:04

I don't know. I, I think you're, you are over-indexing on, on the human side of it. I think, though often it is a collaboration, I would say, between like we may get an idea from the AI that we're like, "Oh, that's sort of a fun area to explore," and then the humans come up with a better result. but I think you are downplaying how often the AI comes up with at least a starting point that becomes fairly relevant to what becomes the eventual title.

Ben Whitelaw 6:32

I think we're winning. I don't, I don't think we're winning.

Mike Masnick 6:34

I, I think humans are winning. I think if we're going on a straight like first pass to post 50/50 sort of situation, th- there is more human, elements to our titles than AI. But, I don't think 90/10 is the, uh, is the real breakdown.

Ben Whitelaw 6:49

that's my goal to, uh, to take an un- unassailable lead over a- our AI, uh, creators. whether you like human-created titles or AI-inspired ones, we have a Patreon tier for you here at Ctrl Alt Speech. Patreon is how we kind of fund what we do, how we, are able to dedicate time to reading all the news that we talk about each week, and everything that doesn't make it onto the list for our episodes. Our supporters pay $10 a month and get earlier episodes, plus the extended ones. And our, kind of max supporters, the ones that pay $20, they get the chance to suggest stories that we cover each week, plus some ideas, for future episodes, of which we've got a couple we're working on right now. So if you're liking the podcast, consider becoming a Patreon member, patreon.com/ctrlaltspeech. Mike, there's a lot of, a lot of news this week as ever. We have a mix of kind of regulatory news and some kind of AI news. We've, we've been covering more of the latter, I'd say, in the last couple of weeks because of, just the nature of the news agenda. We're gonna tackle a lot of that in our Patreon episode But we're gonna start, I think, with the much touted, much teased EU social media ban approach, which we've touched on, on the podcast before. We've kind of got a sense of what's coming because there was this special panel that the EU Commission convened and which wrote a report that came out in July, and that advocated for a, tiered approach for teenagers that, graduated them through access to social media and different features as they got older. And, and the big announcement this week, or rather lack of one, is the fact that they've, gone with that approach. the headline is that we're seeing under 13s not be allowed a- on any social media access, unless they're with a parent or unless it's a kid-safe app like YouTube Kids, that 13 to 15-year-olds will be allowed kind of additional access. and that they'll be, have what they call a kind of limited account, to be able to kind of graduate through as, as they learn the internet more. And then the over 15s is, where you get certainly more access to the internet, and we'll talk a bit more about what that looks like. this was all announced, Mike, in a big flashy, kind of State of the Union address by Ursula von der Leyen, the European Commission president. And it's almost like not a story because it's been so talked about, but it still merits discussion, I think, in the podcast today.

Mike Masnick 9:04

Yeah, I mean, it, it's definitely a step forward in terms of actually making this a reality. and so I, it is a big story. I think, especially because it will, assuming it goes into effect, it will impact much of Europe. you know, we've talked about France pushing their own social media ban for teens. and like a day before this announcement came out, Macron announced that they were, you know, they had revamped the plan that had been thrown out by the constitutional review. and this would effectively wipe that out. There's a little bit of, you know, it's kind of funny to see him rushing to get this thing done then when it's not gonna matter. but yeah, I mean, this would obviously impact a huge number of people, and so I think it is a, big story. and we'll see. And, and, you know, the thing that struck me about it is that I think there are problems with it. I think that it requires age assurance, as they say, which is age verification, which in the end becomes identity verification. And as I've raised over and over again, there are really big privacy concerns with that, especially when it's mandated. you're going to have bad implementations. You're going to have privacy violated in some form or another, and it's going to put people at risk. So I think that part of it is really dangerous. what I do think is interesting is the sort of graduated approach, which is somewhat different than what we've seen in almost every other one of these, attempts at, banning kids from social media in some form or another, where you have this, under 13 you have this approach. And then the, the more interesting part to me is sort of like that, the 13 to 15-year-old part where they sort of have this idea that you'll have, you know, sort of training wheels for kids getting ready to have, more access to the world, where, you can have access to apps that are supervised by your guardian in some form or another, and they have limited features and limited screen time and, and all this stuff, and it's sort of trying to get people ready. And, if we have to go down this path, and I still think this isn't a path that we should be going down, it is kind of interesting that they, are allowing for a gradual understanding of these apps. Because my fear with the pure bans was this idea that you keep people off these apps entirely and then suddenly they turn whichever age, 16 or 15 or whatever, and suddenly they have full access and they've had no training on it, and they have no idea how to use these things properly, and I don't find that to be particularly safe. so there is something nice in having this, this idea that, okay, you'll, you get sort of like limited access and you sort of learn how to use it appropriately, and then we sort of take off the training wheels, as you get older. So I, I appreciate that aspect of it. It, feels like, for all that I complain about the European regulatory approaches to things, They are a little bit more thoughtful about how they do these things than most other places where it's just sort of, someone comes up with an idea and that's, that's what we're going with.

Ben Whitelaw 11:53

Yeah. And, and credit to them for that, I think. You know, to be honest, the real story was when this special panel report came out in July, which, advocate for this, tiered approach, and which at the time, coming couple of months out of the UK's own social media ban announcement and, yeah, with the Australia implementation kind of running along in the background and going fairly badly according to, a lot of, uh, I guess, metrics. This felt very kind of novel at the time. It felt a bit strange. It felt new. And that, I think, is, uh, that is gonna change over time. I think we're going to see other countries follow suit with what the EU have outlined here with a, year-by-year, a-approach in which people get access to features that, suit them for their age. And, and I think that's a good thing. Um, There's already been talk in the UK of, of the social media ban kind of moving in that direction as a result of this. So the timing is really important. the fact that this has come out now is likely to impact not only the UK's approach, but other, other countries' approaches as well. It's also worth noting, I think, that, Ursula von der Leyen, as the European Commission president, was, was under a lot of pressure to get it out when she did. as you say, France was kind of unhappy with the speed at which the commission was coming up with these ideas. Other countries like Denmark and Greece similar. And so, you know, credit where credit's due, they've come up with a sensible plan for a, a social media ban, if there is such a thing. And it's, y- robust enough that those countries are, are gonna, be satisfied, I think, by the, the route that taking.

Mike Masnick 13:20

Yeah. I mean, the thing that we'll have to see is, again, like I still feel that a lot of these approaches are not based on science and so we don't really know what impact it's going to have. We don't know how helpful it's gonna be. There are a bunch of other features in here, you know, the banning addictive features, which we still don't really know if they're addictive or sort of how that works or what, what does that mean in practice. There are a bunch of questions about this. I still obviously, you just worry about the idea of, you know, as soon as you do age assurance for kids, you're really doing age, assurance for adults, which turns into ID verification in some form or another. so you know, I wouldn't call this perfect, uh, by any means. I have a lot of problems with it. But, I do find it interesting that they're at least taking into account the idea that the pure ban and then suddenly free reign approach is not gonna work. and so, you know, we'll see what happens with it.

Ben Whitelaw 14:14

Yeah. The, other thing to note, which I think is interesting to a lot of our listeners, is the foregrounding of kind of product design obligations within the Kids Act. And this is something we've seen in, in lots of other forms, right? It's, safety by design and, and the product design is where, you know, it's the basis on which Meta was successfully, not, successfully sued, where they settled in the recent case that we talked about in our special episode of Ctrl-Alt-Speech. and it's, seemingly is a kind pain point that, consumers are really focusing on as far as getting, better safety features implemented into products. And the EU Kids Act is no different. You know, there's a real focus on, what kinds of features platforms should have in whether that's logged in or logged out, and at different age groups as well. And so, you things like infinite scrolling and particular types of notifications that drag people back into apps, they're being called out as, violative of the act, which I think is a really interesting route and really solidifies this idea that product design is, is where this kind of battleground is now.

Mike Masnick 15:16

Yeah. And again, this is where I'm, I'm less confident that, the, you know, the belief that these things are necessarily problematic. We don't necessarily know that. What I do think may be helpful about this, I have a lot of problems with the sort of framing of like mandatory safety by design, because that often feels to me like an excuse to blame the company for anything bad that happens while a person is using the technology. The problem when you have tools that are used by, you know, half the world's population is that sometimes bad things happen to those people, and sometimes it happens immediately following them using one of these tools in some form or another, and how much responsibility you put on the person versus the tool, it's very easy to blame the tool, and that's what we, we see a lot. That's what I'm often complaining about. the good side of this, though, is that when you have some element of sort of regulatory demands for safety by design, is that for the people within these companies that actually are working on safety stuff, it does give them more leverage, right? One of the problems that we've talked about and that we see all the time is that these companies all do hire good, thoughtful safety people, but often it becomes a question of priorities among the executives there and You know, when it comes down to it, are they going to invest in real safety, or are they going to do the thing that makes them more money? And Wall Street demands that they make more money, and therefore, the safety side of things often gets sort of pushed down a little bit, which upsets the people who are working on safety. So for the people within these companies that actually do believe in building safer, better products for their users, this gives them some, some leverage in which rather than, you know, when they go talk to Mark Zuckerberg or whoever and they have to justify what they're doing, they can't just make the case based on like, people will be safer and maybe there'll be fewer headlines about how you're killing people. Um, but now they also have like the legal side of it, to argue as well. and so, you know, I've certainly heard from some trust and safety people who appreciate these approaches, but also some who, who worry about it, where you know, once there's a legal requirement for it, as we've again talked about, then suddenly you lose some of the creativity in terms of the approaches that you can take in trying to build safer products because now it becomes a compliance discussion. And so rather than just having safety people in the room experimenting and testing things out and figuring out what actually makes people safer, you also have lawyers in the room, and the lawyers are often determining what is allowed and what is not allowed, and that may not be the safest thing. That might be, you know, the lawyers are involved, it's the thing that will give you the least liability is generally the way it works. and that is a very different approach, which might not be the safest approach. So there's, again, it's like a lot of, some good, some bad. what the actual impact will be long term, we still don't know. I think the assumption that, laws like this will suddenly magically make these platforms legitimately safer, is wishful thinking. I don't think that's what will turn up in the end.

Ben Whitelaw 18:13

No, I think that's fair. I mean, the one kind of mini surprise that, I guess came out in the announcement is that not only does the Kids Act apply to social media platforms and also AI chatbots, which is one of the recommendations of, of the July report, but that it also applies to video game platforms that, quote, pose a, a risk to minors," which again is like, not really sure where we draw the line there. But what did you m- what did you make of, of the fact that kind of online games have been included in this? And there has been some interesting kind of chatter in, games industry circles noting that this is, this is now gonna apply to games providers and, studios

Mike Masnick 18:50

Yeah, I mean, we'll also, we'll have to see how that works out in practice. I mean, there is a long history in the video game space of it often being accused of, you know, leading to bad outcomes for kids, and to some extent this feels like a backdoor into re-litigating that. I mean, I think a that fight that happened in sort of the early 2000s about, oh, you know, these video games are bad for kids, and for the most part, that sort of went away when there were a bunch of studies that effectively said n- no, there wasn't any sort of clear negative impact on children playing different kinds of video games. so it feels like this might be a little bit of a backdooring there. but it also, again, points to the fact that like, you know, all of these platforms are different in certain ways, and people interact with them in different ways, and different people have different reactions to them. And I think it's really difficult to sort of put in place a set rules that apply broadly across the board for platforms that are very, very different. and so we'll have to see how it actually works in practice. But, it is interesting to sort of loop that in, 'cause most of the other, teen bans that we've seen elsewhere haven't really gotten to the, the video game and the video game platform side of things.

Ben Whitelaw 19:54

No, they focused really on the big social platforms, better or for worse. Uh, but from my experience, I mean, video games platforms often have the most kind of thoughtful approach to safety and have really the, longest history of, safety work and dealing with different types of harms in very different kind of types of ways. So,

Mike Masnick 20:11

I mean, it depends. I mean, I think you have examples in, in both directions. I think you have examples of, of gaming platforms that have taken it really seriously, and you could come up with some examples of the opposite, so I, I don't... I mean, it is an interesting space. It is one that obviously lots of kids spend a lot of time on different gaming platforms. and so you know, again, like it's interesting to see the experiments. I'm a little skeptical of how successful some of them will be, but, yeah, it's worth noting that they included it in this plan.

Ben Whitelaw 20:40

Yeah. Okay. so we'll, we'll, look forward to the implementation of, that law. I'm sure there'll be lots of news coming out of it, uh,

Mike Masnick 20:47

yeah. There will be plenty more to talk about as, as those, uh, that gets solidified

Ben Whitelaw 20:53

Yeah, and we'll include, a link to the actual act, text which came out today. so this is very, very fresh. Um, talking about children, in Europe, Mike, I wanted to, in our eye-catching section, talk a bit about school results, from a, test that has been-- is run every three years called PISA. I didn't know that this existed, but This test happens every three years and, and looks at maths and reading results of 15-year-olds. and it, kind of benchmarks them as to, their progress in those regards. And what the latest PISA test found was, I think, something that a lot of people who, care about the Kids Act will also be very interested in. the announcement of the report found basically that 15-year-olds now have the same kind of reading and, maths results as a 14-year-old in 2018. So they've essentially, lost a, a year's worth of, capabilities in the course of the last eight years, which is obviously a massive drop. the PISA survey looks at it in terms of percentage points, but, a very worrying trend in a lot of different countries. And the, re-authors of the report, who are, the OECD, a big public policy forum, put this down to what they call digital distractions. So a combination of short-form videos and latterly AI kind of eroding the ability of children to, think critically and, do work. and that's interesting in of it, in and of itself. A lot of people think, "Yeah, that makes sense. I can see how that would, work." I... a nice narrative to kind of buy into. But a follow-up op-ed from a journalist at The FT" called John Burn-Murdoch finds a slightly different angle to it, and he says that there's a really strong distinction between countries that have curriculum education systems that are very kind of flexible and outcome-based, Finland being an example, Wales, Scotland, and Ireland being more recent advocates of that type of system, and the kind of stricter, more knowledge-based curriculum that, are kind of owned by countries like England and Germany and others. And he looks at the data of the two and, and finds that actually the numbers for the stricter, more knowledge-based systems hold up much better compared to these kind of outcomes-based countries who have reduced the number of tests and also got rid of league tables, which obviously kinda hold schools to account. And I just found the whole thing really fascinating. you know, it's this idea that, you know, an idea seems very plausible in the outset. You know, somebody makes a characterization, the digital distraction is the cause of, declining test results, and then you look into it a bit deeper and there's more to it than it seems.

Mike Masnick 23:22

Yeah. and this goes back to the same point that I feel like we make all the time which is that these things are complicated. Human beings are complicated. the single answer to anything is probably wrong, right? Like, there's a lot of nuance here, and everybody's different, and, it's easy to chalk things up. And we do this all... You know, there's all these historical examples where it's like, "Oh, you know, our children are failing because of X, and therefore we have to change X." And then we look and then, oh, the original metric for measuring this wasn't quite right, and therefore the big change did the wrong thing, and now we're having worse outcomes. You know, I think it is really complicated. It is really interesting, obviously, to look at this data, and there are all sorts of questions like how much did, the COVID shutdowns impact all this stuff? We're still within the range of, you know, students who had some element of that, and you could blame some of it on that. It's always easy to blame it on the new technology. and the reality is it's complicated. We still don't know how to, educate people all that well. We don't know what works, and we don't know, how things work for different kids too, right? Like, there are different populations of kids, and they're taught in different ways. And so, I think to me, the whole thing is like we're still learning. I had an interesting experience and, and I'm not sure how this will go, but, in meeting, my child's high school math teacher was explaining part of the curriculum that he's trying to teach this year, it's a calculus class, and was saying like one of the things that he's actually trying to teach is how to learn in an age where every answer is at your fingertips. And I thought that was a really interesting thing to say, and he's, was, was saying like no adult had that experience. we all had to struggle to figure out answers to certain things. you couldn't just ask and get an answer, and now we can. And what does that mean for education? And he's actually trying to... He said like part of his curriculum is to try and teach kids how to learn in that environment and like the value of learning. And I've, I'm working on another project, which I'll be talking about in probably a month or so, about kids and, and AI and learning about their attitudes and, and seeing a lot of the data about how kids' attitudes towards AI is changing really rapidly from year to year. I think a lot of this is just there's a lot of newness going on, and we're, we're all trying to figure it out. I tend to have faith that over time we figure it out. but I also, know from historical examples that the first reaction that people have, the sort of knee-jerk, "Oh, this is terrible. We must stop this. We must change this," is often wrong in the long run. but people are resilient. People do tend to figure these things out. So I'm, I'm hopeful, but I do think that the results of this test are interesting in that it's sort of a sign that, there are some problems somewhere, but it's a little bit difficult to figure out exactly what they are or how to fix them at this point.

Ben Whitelaw 26:07

Yeah. I totally agree, and I think there's that importance of seeing it in a kind of longer-term trajectory. It was also borne out in this piece which, again, I'm familiar with some of kind of Finland's public policy work, and we've talked about some of its digital, e-government setup in the past on the podcast. But I didn't realize that, you know, one of the reasons why it was, you know, held up in the kind of early 2000s as such a kind of pinnacle for, for education and for, media literacy, they think it's because of the, the hangover from its very strict means of educating kids in the kind of '80s and '90s. And so, was able to kind of carry that on into the 2000s when people went and, looked around it and, and asked, you know, "Why is it working so well?" No one could, could obviously kind of weight that work that it'd been doing as a, as a nation and as an education system. So you have to see it in this kind of longer trajectory, I think, and this piece really got me thinking about, how short-termist we are in some of what we think and what we believe.

Mike Masnick 27:01

Yeah. and just like a recognition of how many different variables play into all of this, and it's impossible. It's so easy to try and pick out, you know, one or two different variables that have an impact, and the reality is that it's, you know, it's an infinite number of different variables that each play into it in different ways, and it's impossible to sort of calculate out. it doesn't mean we shouldn't be trying to improve, and we shouldn't be looking at, declining test scores or whatever and say like, "Oh, that's a concern." But, you know, I don't think we should rush to the, "Oh, it must be this one thing. If we only change this one thing, we'll fix it."

Ben Whitelaw 27:31

Yeah, I think that's true. definitely one to read, listeners. It's, it's a cracking piece of analysis. One thing just to note before we move into our eye-catching, second eye-catching story, Mike, is, a fun story, that we'll add to the show notes. We did a big episode about the Meta settlement, a couple of weeks ago in the US and Kind of fun story that we'll just add is, that TikTok is, is rejecting Meta's ads that are advocating for TikTok the platform to join the safety settlement, which I think is, you know, is, is the most petty thing you could ever hear from a platform, I think.

Mike Masnick 28:05

there's not much to comment on here, but, like, it's petty in both sides, to be honest with you, right? Because it's like, you know, Meta sort of feels forced to advocate for other platforms to accept these same terms. And so it's kind of petty that then they went and tried to take ads on TikTok pushing this settlement that, like, TikTok should join the settlement, and then it's kind of petty that TikTok was like: "No, we're not gonna run your ads." So I, I just found this to be a sort of very funny story, but, you know, I don't think it's worth going too deep on

Ben Whitelaw 28:35

no. no. Critics of, major platforms advertising content policies will, will wonder how they were able to kind of so quickly spot Meta ads but not, not spot other kinds of ads. But we won't go into that. We won't go into that. So yeah, we'll, we'll move now, Mike, onto our, second eye-catching story, which we'll go a bit into detail on, and then we'll, move into our Patreon section of the, podcast. And these are, are really kind of looking at AI stories of the past week. we both put this same story into the spreadsheet this week, Mike, which doesn't always happen. We're, we're fairly good at staying in our lanes, but this 404 Media story I think caught both of our eyes just for the, sheer, like, of course this is happening element.

Mike Masnick 29:14

Yeah, this is, they had this story which, got a lot of headlines, a lot of people were talking about. It was about what is referred to as Project Lily, which, are human beings who are reviewing ChatGPT, chats. You know, the, both the input and the output and trying to im-improve them. And, I mean, I love 404 Media. I think they do great work. I think, they break all amazing stories and, and they do wonderful work. This one struck me as a little bit odd I was a little disappointed in that. It, it sort of struck a sort of conspiratorial tone to it that, "Oh my gosh, can you believe that OpenAI is sending some of your chats to human beings to review?" and yeah, I can believe it, and I assume that was happening, and I assume that is happening with basically all of these services because that is part of the process in how they work. and at the same time, it's kind of what you would expect for all the people who are complaining about, oh, you know, somebody, discussed suicidal ideation with one of these chatbots, and somebody should've been alerted and all of these kinds of things. you know, that is a slightly different thing. There, there's one for, like, just improving the inputs and outputs of, of the chats. there's another for, like, responding to potentially negative chats. But, humans being in the loop in some form or another is not a surprise. There are a few multi-billion dollar companies that are built entirely around humans improving AI chats. I guess maybe that is news to some people. you know, Meta famously tried to buy one of the really big companies that all they do is, have humans involved in trying to improve the chats. I think some people interpreted this as, sort of like, "Oh, it's really humans who are doing the responding on AI's behalf," which is not accurate. It's more just sort of review and improvement of the chats. you know, it's interesting. I think maybe for you and I, we spend all these, time in the trust and safety world, and we recognize trust and safety world, and we know that the trust and safety world is often built on human labor, you know, often in, problematic ways and often, you know, low-paid foreign workers. We've discussed them on this podcast many times, that you have, like, what used to be call centers is now people reviewing content. it's not a surprise to me that that is the exact same thing that is happening in the AI space, that there are humans, probably often in relatively low-paid jobs in far-flung places around the world reviewing AI chats in different ways. the only thing that came out of this that I think was really interesting was perhaps the companies could be more forthcoming about that. there was sort of a checkbox in there which is like, "Do you want your AI chats to help improve the system?" That they've now tried to clarify. OpenAI at least has, has made it a little clearer that some of your chats may be shared with other people to review. That wasn't as clear beforehand. So maybe it's a little better on the education front. but this story struck me as like, it's interesting, but it was almost more interesting to me that so many people seem surprised by it than that it existed, that, that this is what was happening.

Ben Whitelaw 32:21

I think this is another new paradigm that we're facing. You know, it's, this idea that in the kind of social media platform world, people knew that their content was being moderated. That was part of their, the kind of Faustian bargain that they'd set up, um, when they were, getting access to major platforms for free. they probably didn't know that they were the product, but they, they knew that there was somebody somewhere that's gonna be looking at some of their inputs, whether it's on a public feed or in a group or whatever. I don't think people have quite cottoned on that, this is the case with the chatbots because of the, the interface, because of the kind of nature of the im- information they input, the very personal information, as you say. And, the reaction I think is, people's eyes being opened to the fact that there's a lot of similarities between the, kind of Web 2.0 platforms and these new, frontier AI companies in terms of what they're having to do to keep users safe. they might differ in other respects, but in terms of what they need to do, there's It's a lot of the same playbook, and if you speak to lots of trust and safety professionals, they'll say, "You know, this is something I did 10 years ago in a slightly different guise." but no one told users that, unfortunately.

Mike Masnick 33:28

Yeah, and I guess, you know, when you're using social media, you know you're sort of posting if not publicly, but to a group of other people that you know will read it. Whereas when you're using AI, it does feel like a very one-on-one situation. So the idea that another human might read it does feel a little bit more invasive in, in some form or another. And so, you know, to that point, I guess it is useful for more people to know about it. it just didn't strike me as, like, this big conspiracy that it was, like, being hidden. But, you know, I, I guess I was wrong in that I assumed more people realized it than maybe they did.

Ben Whitelaw 33:59

Yeah. they just don't listen to the podcast, Mike. That's

Mike Masnick 34:02

Yes. Yeah. Well, this is why more people should listen to the podcast. So if you're listening to this podcast and there are people in your life that didn't know that your AI chats may be read by human reviewers, you should tell them to listen to this podcast

Ben Whitelaw 34:14

Indeed, indeed. you can also go to some of our new social media channels where you can see clips of us talking about these topics. And also send those clips to family, friends, colleagues who might be interested in the podcast. Rate and review us wherever you get your podcasts. It really helps us, reach new audiences and, and reach, people who, who want to know more about how the internet works, but just haven't been exposed, let's say that. Mike, we're gonna wrap up our free episode there. We're gonna move into our extended Patreon supporter edition and talk a bit about the topic that has kind of been on everyone's lips this week, which, which is pacing the frontier. tell us why people should tune into this extended episode considering they've probably read something about what we're about to discuss. What's, gonna be different this time around?

Mike Masnick 34:59

Yeah. I mean, there have been stories for, you know, since AI, burst on the scene six years ago, or four years ago really, about, you know, oh, maybe the, labs need to slow down, and those never go anywhere. And in the last week, we've suddenly had a lot of discussion and seeming agreement among the big AI labs that maybe they actually do need to do something. and it seems to have sort of turned a corner in some areas, and so I think it's worth discussing what does that mean, what does it mean for the safety of these tools, and so that's what we're gonna go into.

Ben Whitelaw 35:35

Great. Let's do it

Announcer 35:38

That's the end of this week's free edition of Ctrl Alt Speech. To hear the extended episode with additional coverage, become a Patreon supporter at patreon.com/ctrlaltspeech. That's C-T-R-L Alt Speech. If your organization is interested in sponsoring the podcast, contact us via our website at ctrlaltspeech.com. Thanks for listening