WEBVTT

1
00:00:00.000 --> 00:00:24.000
The whole mechanism of the game is built on a gambling-style addiction loop — the only point of it is to get you coming back about ten times a day to do something in the game. Inward-turning, self-image-related mental health problems have also been rising among boys. Now, these are typically more common in young girls, but things like body dysmorphia — seeing yourself in the mirror as different from how you really look — for boys that might

2
00:00:24.000 --> 00:00:44.000
mean thinking your biceps are too small — cases like that have been increasing. In practice, the algorithm is estimated to account for about 80% of TikTok's total value. Anyway — welcome to the Neuvottelija channel. I'm on location today, at Isak Rautio's Futucast studio. Great to be here in your studio.

3
00:00:44.000 --> 00:01:04.000
I used to be over at your old place. Welcome to our studio — we're trying out a double episode, meaning we'll release two episodes at roughly the same time, loosely on the same topic: the anxious generation. I don't know if we actually are an anxious generation — Mikko's almost my age — but Mikko Alasaarela is also here as a guest. Hello. Hi.

4
00:01:04.000 --> 00:01:33.000
Nice to be here. You're often up there a lot, though — you started out as Finland's go-to TikTok whisperer, and these days you get to combine that with some AI expertise as well. Yeah, well, this TikTok expertise really comes out of my focus on algorithmic influence, and once TikTok became the eye of the storm, so to speak, for algorithmic influence, it was clear as

5
00:01:33.000 --> 00:01:58.000
day that my expertise and insight were needed there. Right — as our starting point, we've both been studying a book by one of America's most qualified sociologists: Jonathan Haidt's 'The Anxious Generation.' It's been a pretty impactful book, so should we start with some thoughts on it? Mikko, do you want to start with what Jonathan Haidt actually says? Yeah,

6
00:01:58.000 --> 00:02:26.000
so basically, there's long been talk that today's social media model is driving a lot of negative developments among young people — that's been discussed a lot — but researchers have always said in those discussions, 'hey, correlation doesn't equal causation.' And the whole point of Jonathan Haidt's book was to show that there is

7
00:02:26.000 --> 00:02:52.000
also a very clear causal link between social media and smartphone use and young people's ill-being. Yeah — what about you, Isak? You're not quite part of this anxious generation, meaning we're talking about people who weren't actually messed up by COVID, but by the phone. You're a bit younger than us though... so actually, before the episode I was wondering why I'd even been invited here. You're more like an algorithm scientist,

8
00:02:52.000 --> 00:03:22.000
Mikko, and you're the one who's read the book. Same goes for me, the podcast host — my best theory right now is that you still think I'm 25 years old, so I count as part of this anxious generation — I'm basically here as an expert by experience. No, but seriously, I should probably say up front that I don't have any professional qualifications to critique everything Haidt says in the book, but I'm fairly favorably biased toward him, so I'm quite ready to take his message on board. I think he's a good voice, a good addition to today's

9
00:03:22.000 --> 00:03:48.000
discussion culture. This question feels very valid, and I think he's really good at putting these views forward. So, right at the start, before we get into the real substance, it's probably worth saying that I lean — my gut feeling, at least, leans a bit toward Haidt being pretty right about this in some way. And in that sense, it is a scientific book; he's written really good books before too, on this

10
00:03:48.000 --> 00:04:18.000
topic — like his theory on people's political decision-making, really great, sharp analysis. So I think this is very much a science book, and in my opinion, that initial thesis — that the smartphone ruined young people — well, it does sound true in a way, especially compared to when people explain it away as, say, COVID isolation ruining young people. There, you're looking at a much simpler correlation again. I was speaking, a couple of

11
00:04:18.000 --> 00:04:45.000
years ago, at a management seminar for the management teams of Finland's university central hospitals, about the impact of AI on their field, and they told me themselves that demand for adolescent psychiatric services exploded almost exactly a year after smartphones became common among young people. That's also just a correlation whose causation hasn't been scientifically confirmed, but it's not exactly

12
00:04:45.000 --> 00:04:53.000
rocket science to understand that there might be a very concrete impact there. Yes, and that's exactly where you saw it in the data, i.e.

13
00:04:53.000 --> 00:05:18.000
the mental health curves jumped. So at first I thought — I've sometimes been critical, in a way, of this self-diagnosis culture, this kind of 'woke' self-diagnosis thing — but this book, I think, convinced me that it's actually quite true. People have, quite literally, broken down mentally during this smartphone era, so it's not just some kind of self-suggestion.

14
00:05:18.000 --> 00:05:21.000
Should we start drawing out a distinction between Haidt's different theses?

15
00:05:21.000 --> 00:05:48.000
On the one hand, he talks about the harms of social media — that people's, and young people's, lives are on a stage, so a large audience of adults sees what they do and say, and it stays there forever. You can always go back and see what kind of person someone was a few months or years ago, and that creates very specific pressure on young people to be a certain way, which is distressing. And then, on the other hand, as social media use increases, more and more other

16
00:05:48.000 --> 00:06:18.000
childhood habits and ways of growing up get missed. For example, this lack of a play-based childhood — that's maybe not a direct effect of social media, but a side effect of increased social media use, and that too could... I don't know, is that just a question of definition, or is it also a direct harm? Well, on top of that, you could add the point that for boys, this marginalization and addiction are also driven by games, so it's not just

17
00:06:18.000 --> 00:06:44.000
social media causing these problems — we need to remember it's several factors pulling in the same direction. Yeah, I'm slightly of a different opinion there. I mean, I'm a man who's... on an episode with Tuomas Enbuske, we wondered whether we're both somewhere on some kind of neurodivergent spectrum — you could even say on the autism spectrum — and this is a masculine

18
00:06:44.000 --> 00:06:55.000
trait, this craving for structure and collecting, or for systemic detail, and games give us that, so I've...

19
00:06:55.000 --> 00:06:57.000
made a bit of a confession here myself, where I said that I...

20
00:06:57.000 --> 00:07:05.000
played through, for example, Baldur's Gate 3 on Honor Mode, simply because I thought it was a fun, self-contained...

21
00:07:05.000 --> 00:07:08.000
goal that gave me satisfaction, and I didn't feel like my world...

22
00:07:08.000 --> 00:07:13.000
suffered for it in any way. Well, my view on that is that games you can...

23
00:07:13.000 --> 00:07:17.000
actually play through to the end are exactly the kind of games that contribute positively; then...

24
00:07:17.000 --> 00:07:27.000
there are all these other games — meaning 85% of all games — which are free-to-play, never-ending experiences where the whole game...

25
00:07:27.000 --> 00:07:32.000
mechanism is built on a kind of gambling-style addiction loop, and...

26
00:07:32.000 --> 00:08:02.000
its only purpose is to get you coming back about ten times a day to do something in the game. But Isak's point about play, I think, cut a bit deeper than that — there's Taleb's antifragility theory, meaning, in a way, we have an anxious generation because it's a fragile generation, because it hasn't been toughened up. Real, proper, rugged kids who were put out in nature to play, to take some hits, to fail a bit

27
00:08:02.000 --> 00:08:30.000
and to take hits from the world, and to experience disappointment and boredom — all the things these 'curling parents' have since removed from kids' lives. Yes, but let me comment on the earlier point about games — in my experience too, I'm a child of the internet, even though I graduated from high school the very year those hockey-stick curves Haidt talks about, in young girls' mental health, started to show up — meaning 2012, when smartphones and social

28
00:08:30.000 --> 00:08:56.000
media became part of young people's lives — that's the year I graduated. But I still grew up on the internet, and I think I turned out a relatively normal, well-functioning young person, even though a lot of people might say the internet raised me as much as my family did. I don't know if that's entirely true, but, for example, at 13 I saw a video online of a man being murdered with a hammer, because back then there were these gore sites that every teenage boy went

29
00:08:56.000 --> 00:09:23.000
to watch. That really shouldn't have been part of any young person's childhood — seeing something like that. And maybe, just as a side note here, the scary observation is this: of course it was shocking, I still remember it, I've never really forgotten it. I remember how shocking it was even back then to watch — but it probably would have hit far more directly if I'd seen it happen live rather than on video. That's a good point, actually, because Jonathan Haidt says in his book

30
00:09:23.000 --> 00:09:45.000
also that we protect children to the maximum from real-world dangers that might actually be useful for their development, while at the same time we don't monitor their internet use, which can let young people drift into all sorts of rabbit holes — extreme violence,

31
00:09:45.000 --> 00:10:12.000
extreme porn, extreme views, that sort of thing — and into genuinely unsafe spaces too, like Snapchat, which pedophiles use extensively to groom children. And we don't recognize these dangers, because it doesn't look dangerous when someone's just staring at their phone. Right — let's move to the girls' side. In a way, the dangers for girls are partly what

32
00:10:12.000 --> 00:10:21.000
you just described — there's sexual harassment and grooming — but on top of that, social pressure plays out differently for them.

33
00:10:21.000 --> 00:10:49.000
So in a way, boys are partly protected from that, even though boys have their own harsh hierarchies too, which can be used cruelly for bullying — but for girls, it's particularly cruel. So girls growing up on the internet end up facing things in a completely different way — pressure around appearance, and pressure to succeed and get likes, in a way that's different from boys. That's true, and social media specifically has a more harmful effect

34
00:10:49.000 --> 00:11:16.000
on girls than on boys, and it comes specifically through these distorted role models — these mental images that have nothing to do with normal, real life. As a result, you can literally see it on the streets today: young people getting plastic surgery to look more like some Instagram model — that kind of thing

35
00:11:16.000 --> 00:11:44.000
has increased a lot, and probably specifically among girls. Exactly — and that's a harsh world, especially combined with that constant pressure to be accepted 24/7, and, in the social hierarchy, to at least not be bullied, but ideally be the popular girl. So it drives things like anorexia, weight-control problems, and other genuinely unhealthy stuff, in a way,

36
00:11:44.000 --> 00:12:11.000
of constant self-reflection through a negative lens. Without trying to pit girls against boys here, or girls against each other, this is just an observation I've made — I've done some loose research for a possible project related to bodybuilding culture and steroids, and steroid use among young people has increased. There's some vague evidence — I've only read the headlines and abstracts, and maybe only skimmed even those — but it does seem that these... I don't know if this is the right term...

37
00:12:11.000 --> 00:12:40.000
'internalizing' mental health problems — inward-turning, self-image-focused mental health problems — have also increased among boys. They're typically more characteristic of young girls, but things like body dysmorphia — seeing yourself in the mirror as different from how you actually look, thinking you're too small — have also been increasing among boys. Again, I'm not drawing any hard correlation here, this is just an informal observation, but it does seem that way. And Haidt also writes

38
00:12:40.000 --> 00:13:06.000
in his book that social media affects young girls' self-image and mental health in a completely different way than it does boys. One dynamic here — not necessarily something the platforms drive intentionally, but something that simply comes out of algorithmic optimization — is that all our individual worlds drift in some direction, as these platforms try to find us content they think we'll like to watch.

39
00:13:06.000 --> 00:13:33.000
And if you start watching a certain kind of content, just out of interest or curiosity, it can start feeding you so much more of the same that you just drift in that direction because of that original interest. And as a result, you now see, in every Western country, a significant divergence in worldview between girls and boys, based on exactly this: boys get pushed content that's suitable for boys, or seemingly

40
00:13:33.000 --> 00:14:01.000
suitable for boys, and girls get girls' content — and we end up in a situation where it's no longer possible to have a real dialogue between the genders, because their worldviews are seen as so far apart that even heterosexual pairing starts to feel unrealistic. This has already played out concretely in South Korea, where the gap is so large that

41
00:14:01.000 --> 00:14:30.000
having a heterosexual relationship is no longer even the goal for a lot of young people. Yeah, and this divergence in worldview has happened through an algorithmic, interest-based internet. Maybe Mikko could give us a bit of insight into what that term actually means. Yeah, so I actually got into this algorithmic-influence work after I'd spent time in Silicon Valley, right during the explosion of the early phase of social media,

42
00:14:30.000 --> 00:14:46.000
and I was lucky enough to be around for that. Wait, was that photo of you real? Yeah — I actually thought it was a deepfake, because you didn't give any context for it. Yeah, it wasn't a deepfake — I really did hang around with Mark Zuckerberg and others in Silicon Valley. Let's put that photo into the video version of this episode.

43
00:14:46.000 --> 00:15:14.000
Listening to this — yeah, well, a lot of people who knew I'd been out there said, 'yeah, you really were hanging around with them.' Not that many people suspected it was an AI image, because people already knew I'd been in those circles. I thought people assumed it was, like, Mark getting to hang out with Mikko, but yeah — let's just say these kinds of famous

44
00:15:14.000 --> 00:15:41.000
entrepreneurs are a lot more approachable before they're as rich and famous as they are now. For example, I went to dinner with Drew Houston, the founder and CEO of Dropbox, and other well-known entrepreneurs — hung out with Mark Pincus, the founder of Zynga, and others. These well-known entrepreneurs — I'll say it again — were all a bit less famous and a lot more approachable back then than they are today. But the conversations I got to have with these

45
00:15:41.000 --> 00:16:07.000
industry movers and shakers, the people who actually built these big platforms, let me see the change that was happening as it happened, because I was right in the middle of it myself, when social apps and the like button were introduced — added to Facebook — and that's when a wave started where all content on the internet became algorithmically curated. Meaning

46
00:16:07.000 --> 00:16:35.000
before that, you had, say, a Myspace page somewhere, and you'd go and look at that specific page. It was practically like everyone had their own website, and you'd go check now and then whether someone had updated something on it. Then the algorithmic feed came along, and content was selected for you based on what might interest you, with the like button working as the curation mechanism for that content. And this wasn't the only trend happening — similar things were happening

47
00:16:35.000 --> 00:16:48.000
across every area of digital content. We went from buying a one-dollar song or a ten-dollar album on iTunes to a Spotify-style 'all-you-can-eat' model.

48
00:16:48.000 --> 00:17:18.000
we moved from video rental to Netflix, we moved from buying an Angry Birds-type game to Supercell's model — to that free-to-play world designed to keep you addicted to playing for as long as possible. Some might point out that someone who played through Baldur's Gate 3 for a one-time fee wasn't affected the same way — it's the same thing, right? Sorry, but to continue: I'd say that Angry Birds world, where you just bought a game outright, was so much healthier,

49
00:17:18.000 --> 00:17:46.000
for young people, because that gaming experience always has a beginning and an end — the goal in practice isn't to keep you hooked forever. And all these trends happened around that same time, in a fast cycle after 2010. It was around 2009 that the like button arrived and things started shifting in this direction. Social games first appeared on top of Facebook, then moved to mobile, and once they were on mobile, that spread to young people — so around 2012–13 this

50
00:17:46.000 --> 00:18:15.000
smartphone penetration among young people started growing explosively. And you see these massive shifts, like the gaming industry moving from buy-to-own to free-to-play within a single year — 2011. At the start of that year, free-to-play made up 30 percent of games; by the end of the year, it was 70 percent. Today it's 85% — meaning 85% of all games out

51
00:18:15.000 --> 00:18:19.000
there are basically free-to-play, addictive games. I remember...

52
00:18:19.000 --> 00:18:46.000
from my own experience, when Facebook first came out and we logged in — as I remember it was '08, or maybe '07 — and, well, back then it felt much more like a... what's the word for 'supplement' — like an add-on. It was a way to live your high school reality a bit better, but it was still the same old high school reality we had

53
00:18:46.000 --> 00:19:12.000
where you could find out, say, what the homework was for the next day — if we ever even talked about that — who was single, who wasn't, where things were happening in the real world. It didn't suck you into some new reality; instead, people went there to get more information about the actual reality. A lightly augmented reality, yes. Yeah, yeah, yeah — and this development, once it started back then, it

54
00:19:12.000 --> 00:19:41.000
kept moving forward over the years, so that after that first wave — after the Facebook like button had arrived — came follower-based social media, a shift away from friend-based social media, which back then was about creating a two-way connection where both people see each other's content. We moved into following influencers, and that's when the influencer market was born — Instagram arrived, Twitter arrived, plenty of other follower-based platforms came along, YouTube stars

55
00:19:41.000 --> 00:19:53.000
the podcast — so basically you can still follow that too. Quick side note here: Isak and I would still really appreciate it if you subscribed to both Futucast and Neuvottelija — it warms our hearts, and the algorithm a bit too.

56
00:19:53.000 --> 00:20:23.000
Exactly. Yes, on YouTube, follower counts still matter a fair bit — a big deal, actually — so it's genuinely worth subscribing, it makes a real difference for us. On TikTok, apparently, follower counts no longer have much impact on a video's reach at all — it's moved purely into algorithmic social media. So basically, follower-based social media changed the dynamic to be less social, because you were no longer following your friends — so it

57
00:20:23.000 --> 00:20:52.000
stopped being just a place to find friends, and became a place to seek out content from entertaining, interesting people — celebrities and others producing entertainment for you at a steady pace. And then the next wave after follower-based social media was algorithmic social media, the TikTok world. And in algorithmic social media, even that got stripped away — it was basically declared that it doesn't matter who you follow, we'll show you

58
00:20:52.000 --> 00:21:20.000
the most addictive videos possible, out of all the content available on the entire platform, in vertical video format, at such a pace that every 15 seconds you get a new video, in an endless feed. It'll never end — you can watch it as long as you want, you could spend your entire life on it and it still wouldn't run out — that endless chain of videos. And TikTok started

59
00:21:20.000 --> 00:21:45.000
rising rapidly, and eventually TikTok's rise forced all the other social media platforms to jump on the same train, because TikTok was clearly beating them on the time-spent metric — and time spent translates directly into money through ad revenue. So Instagram joined the race by changing their

60
00:21:45.000 --> 00:22:12.000
content to be more algorithmic. And this new Threads platform from Meta is purely based on a 'for you' algorithm — there's technically a following feed option, but I doubt many people actually use it. And YouTube is now roughly 80% driven by that algorithmic recommendation mechanism, so following still has a small impact.

61
00:22:12.000 --> 00:22:31.000
And the algorithm naturally recommends channels to you more if you follow them. But X, too, shifted purely to a 'for you' algorithm after Elon bought it, and it has actually made the feed a bit more addictive — even though it's absolutely full of filth.

62
00:22:31.000 --> 00:23:00.000
The thing is, I don't really know why, but you get hooked on some pretty strange stuff, like fights — and I get something like twenty thousand impressions just from mentioning Finland's high taxes. We compared notes on our own For You pages — I don't know if it's Elon's algorithm being that personally tailored to your preferences, but it's a bit like getting hooked on some B-movie where unexpected twists keep happening

63
00:23:00.000 --> 00:23:12.000
Yeah, but how much — I mean TikTok was originally Musical.ly, and it started out with dance videos and this use of soundtracks — how much do you think that's...

64
00:23:12.000 --> 00:23:42.000
music, and the vertical, short format itself, that drive it? Are those the three most important sources of addictiveness for that format, in a way? Well, the best thing about the short-video format was that it let the algorithm learn a massive amount about your own behavior very quickly, because when videos are that short, you can get through hundreds of them in just a few hours — and based on that, you can

65
00:23:42.000 --> 00:24:10.000
build a really good predictive model of your behavior. As a result, this model has taken over, and of course, once you're in this kind of fast-paced cycle, people's brains start adapting to that pace, so we end up with a kind of goldfish attention span — our patience decreases, our ability to focus weakens. But our focus is still good enough, though, to stay on those kinds of

66
00:24:10.000 --> 00:24:37.000
platforms — so your focus basically atrophies down to just that level. Yeah, but the point is you have to be constantly fed new things like this to stay hooked, because if you get something too long or boring, you just close it and do something else. About the music — I actually sometimes watch the Instagram feed with the sound off, but, in a way, it probably mattered a lot at first — those music dance challenges — and that challenge aspect has been part of it, and

67
00:24:37.000 --> 00:25:06.000
in a way, some unifying genre, whether that's music or something else — so it's probably still at the core; that is the world it started from, after all. But I'd say, yeah, TikTok is still used a lot for music distribution — TikTok actually makes deals with record labels. They've built a network of influencers where the influencers agree to promote whatever TikTok asks, in exchange for getting an algorithmic boost for their own videos.

68
00:25:06.000 --> 00:25:32.000
And since they've got that influencer network built in behind it, they can basically push any song to number one on the charts, purely by feeding people TikTok videos that feature that song. We could probably go on about that forever — let's move over to the political side. So, according to one theory, this is essentially a culture war, the East trying to destroy the West with a psyop

69
00:25:32.000 --> 00:25:59.000
so this is part of a campaign to push as much destructive, self-destructive content as possible, and in that way we get ruined. Well, it's a fairly easy conclusion to draw, because TikTok's content is actually banned in China, and yet TikTok's content is curated in China — a pretty strange equation, that Chinese censors are going through and editing that content, while it's not shown to their own people at all,

70
00:25:59.000 --> 00:26:26.000
and then a local version has been made whose content is clearly a lot healthier than what's shown in the West. It's pretty hard to argue there's no intent behind that, and this is exactly what makes it difficult to regulate, because when Trump first tried to ban TikTok in the US, in 2020, a judge from Pennsylvania

71
00:26:26.000 --> 00:26:47.000
rejected those two orders — I don't know the Finnish term for them — and Montana, as one state, decided that if the federal government won't deal with it, if the president won't deal with this, then they'd handle it themselves and ban it on their own. They also pointed to the fact that TikTok encourages young people to do all sorts of crazy challenges with real-life consequences — things like the 'let's go trip...'

72
00:26:47.000 --> 00:27:16.000
challenge, where some random person on the street might hit someone on the head and cause real harm, and all sorts of other things. The record apparently mentioned something about smearing feces on children, or other crazy stuff like that. But anyway — one argument from the court, there were three points, but one was specifically that it can't just be that these are examples of expression you don't like, so you decide to call them harmful — but

73
00:27:16.000 --> 00:27:44.000
they're nevertheless examples of expression that's constitutionally protected in the US. You can't cherry-pick public discourse by saying 'we like this' and 'we don't like that' — meaning, without other grounds, simply not liking something isn't a sufficient legal criterion, even where influence operations are involved. The influence angle is then a separate, national-security-related argument the federal government can make if it wants to pass a law banning TikTok — a separate discussion, but one that has indeed come up, under Biden

74
00:27:44.000 --> 00:28:10.000
but this is tricky precisely because for us, this expression is essentially free. But there's also this other side — there's been a lot of debate in the States about whether algorithmic feeds are, in practice, editorial curation decisions made by these platforms, because after all, traditional media is regulated precisely because it chooses what news to show. So why should social media platforms get off scot-free if, in practice, they

75
00:28:10.000 --> 00:28:39.000
are algorithmically deciding what you see — Title 49, Section 230 of the Communications Decency Act, from '97 I think, is the section that defines that a third-party platform like this can't be treated as a content publisher, only as a content intermediary. And this is exactly the law that's now been changed in Europe — through GDPR or the DSA, I think — a change has come here, so you, or the Commission, can now challenge, for example,

76
00:28:39.000 --> 00:29:09.000
TikTok — like they're doing now — over whether, under GDPR, TikTok has sent data to China that it shouldn't have. I think there's a report coming out of Ireland this summer on that, they'll publish their findings on whether TikTok has misrepresented this. But now we — meaning the EU — can take these platforms to court if we think they're doing something illegal. Yeah, but that algorithm point is interesting — I was listening to... sorry, on illegal expression —

77
00:29:09.000 --> 00:29:38.000
so Merja Ylä-Anttila, the CEO of Yle — Finland's national public broadcaster — basically got interviewed by a sympathetic Yle journalist, and there was this funny quote where she said, 'we tried this,' since Yle has also started collecting user data by requiring logins on at least some of its platforms. She said, 'we tried it — Netflix-style, a free algorithm — and it started serving up way too much entertainment, so now we have this better algorithm that gives Finns education and culture instead.'

78
00:29:38.000 --> 00:30:06.000
I got a bit of a feeling of, well, should that even have been published — it's pretty rare to hear that kind of thing from Yle. But I'd genuinely like to know how your public-education mission actually works in practice — is it politically steered in some way? Because this always happens: someone gets to define how the public is educated, or corrupted, and you can respond to that thought. But then the second thing is that we adapt ourselves to the algorithm too. We talked about this

79
00:30:06.000 --> 00:30:19.000
with Mikko in previous episodes — that even I, in this very episode, am partly... I'm not exactly censoring myself, but I am steering the angle of this discussion in an algorithm-friendly direction. So, two thoughts, or...

80
00:30:19.000 --> 00:30:48.000
someone else decides on your behalf what gets shown, and then you, as a content producer, decide to adapt to that. In a Venn diagram, these probably aren't two separate circles, because they likely adapt to each other in some way. You probably do have some interest in building an audience, even though in our world we talk about this term 'audience capture.' It's maybe related to this in some way, except there the external party isn't a platform, but the audience itself, giving you that validation, or financial compensation, or whatever else you want

81
00:30:48.000 --> 00:31:00.000
from there — your clicks and volume. Right, so that's when we get into the real grind: if you want clicks and volume, fame and exposure, then you start playing by the algorithm's rules.

82
00:31:00.000 --> 00:31:29.000
And that's a good point — we tend to think it's the users on these platforms who get addicted to the algorithm's content, but that same algorithm actually gets the creators hooked too. In practice, what happens is that to stay in the algorithm's good graces — to keep your content visible on TikTok, for example — you have to constantly pump out new content that hits the algorithm just right. If

83
00:31:29.000 --> 00:31:56.000
you leave a gap, your ranking and visibility in that whole system immediately drop. And that's led to a situation where there are millions of content creators worldwide with mental health problems, because they're caught in this content-creation hook they can't escape. That's also one of those problems that doesn't get talked about as much, because people assume the real victims are the consumers, and not so much

84
00:31:56.000 --> 00:32:24.000
the creators — but yes, indeed, very negative things also happen to the creators as a result of this algorithm. Now, if we take Isak's Venn diagrams and try to combine them — that's roughly the solution being pursued in the US: this law says the algorithm should be taken away from Chinese ownership and handed over to purely American ownership, where it could then potentially be determined what counts as an approved algorithm. And then, say, from Jonathan Haidt's

85
00:32:24.000 --> 00:32:50.000
antifragility toolbox, you could remove the addictive junk, the dark patterns, the addictive rabbit holes. Could that be one solution? TikTok gave its answer to this — it's been proposed. This is what Biden has now said: sell, or you'll be banned. And that sale scenario would likely mean it ends up as some US company — or some other suitably boring one, sure — I mean, in a way it could be any other country

86
00:32:50.000 --> 00:33:03.000
other than Iran, China, Russia, Venezuela, or North Korea that ends up buying the company, or some entity from one of those places. But most likely it'd be an American buyer if this actually happened. TikTok itself responded to this, I believe.

87
00:33:03.000 --> 00:33:30.000
saying, well, of course they consider it fundamentally wrong, and they have interests to defend here. The argument was that this recommendation algorithm is such important, secret business information for them that it wouldn't be legally, technically, or politically possible to hand it over — especially not within the time window they were given.

88
00:33:30.000 --> 00:33:58.000
And then I believe the Chinese Communist Party also said they won't agree to this — they actually passed a law in China against it, saying that TikTok's viewing data, meaning the data on what people watch, plus the algorithm itself, whatever it is, cannot be taken out of the country. And it's pretty clear they're also covering themselves here, because if that data went over, it would be possible to investigate everything that's been done with it, and I don't think China wants that investigated. So if someone

89
00:33:58.000 --> 00:34:25.000
buys it, someone basically has to train the algorithm from scratch, and the algorithm is estimated to account for about 80% of TikTok's entire value. So the company's value without it isn't necessarily anywhere near as high — meaning it's entirely possible that if someone buys it, its popularity collapses as a direct result. Yes, but there's also that political friction at play, and I recall seeing somewhere that

90
00:34:25.000 --> 00:34:52.000
compared to Instagram, for example, there's roughly ten times more Israel-Hamas-related meme content aimed at girls than there is on Instagram — as if there's a fairly deliberate attempt to mess with girls, or at least radicalize them toward a political angle that's useful to China. Naturally, nobody wants that. And then, of course, the Uighur situation is a total non-starter — that gets blocked out entirely

91
00:34:52.000 --> 00:35:20.000
so yeah, but then again, a US court said this isn't okay — you can't just ban something based on restricting a message. It's an unconstitutional argument if you're targeting some specific message, like 'hey, there's communism here' or whatever. Or else this platform exemption gets removed in the US too — one that, well, has never really existed in Europe anyway, this platform protection

92
00:35:20.000 --> 00:35:24.000
for freedom of speech. So in practice, the US can't even very easily...

93
00:35:24.000 --> 00:35:29.000
ban it purely on grounds of addiction or harm to young people, even...

94
00:35:29.000 --> 00:35:52.000
because in practice they'd then have to ban Instagram and YouTube too. But they can raise the national-security angle, because there's a crystal-clear argument there: from their perspective, you basically have a hostile state manipulating the algorithm — there's enough evidence for everyone that this has happened — which also gives it the chance to spy on everything at the same time.

95
00:35:52.000 --> 00:35:55.000
What happens in the US affects the entire conversation. What happens in the US

96
00:35:55.000 --> 00:36:23.000
carries weight — yes, that's true. So now that the federal government is forced to make those arguments again, updated versions of the 2020 arguments made under Trump, it may well reveal something new. They may also have deliberately held back some things, not wanting to reveal absolutely everything China's capable of doing with this. And maybe back in 2020, it wasn't fully understood yet what kind of effect this could have. But yeah, I wouldn't argue against it either — I watched a Legal Eagle video, which is a really

97
00:36:23.000 --> 00:36:30.000
good channel actually, does good public-education content like this, and I thought it was a pretty solid assessment of all this, even though — well, he's a lawyer, so...

98
00:36:30.000 --> 00:36:57.000
he covers legal issues around the US, and he said that he wouldn't want to argue against the US government on national security grounds — national security is such a strong argument for the federal government at that level that he wouldn't push back against it. But, well — go ahead — if we think about it: what if, say, Russia set up some kind of propaganda TV channel whose job was

99
00:36:57.000 --> 00:37:23.000
to brainwash the population of some country toward a certain stance — that would get banned pretty easily if you were at war with an actor like that. So, in practice, with those same arguments, you can ban this kind of, in quotes, 'social media.' I actually call TikTok, and platforms like it that use algorithmic feeds as their main distribution channel, 'anti-social media,' because it reduces social interaction rather than increasing it, and it

100
00:37:23.000 --> 00:37:53.000
shifts your time and attention away from social interaction toward consuming entertainment. And if you look, for example, at that Gallup study from last autumn on American teens' time use, teens spend 4.8 hours a day on average on this kind of social media, of which 4.3 hours are spent on TikTok, YouTube, and Instagram. So if that time is basically spent mainly watching short videos, then indeed,

101
00:37:53.000 --> 00:37:56.000
that's a good few hundred short videos per person, per day, being consumed there.

102
00:37:56.000 --> 00:38:01.000
as teenagers. And then you have to think about what's being fed into that, whether...

103
00:38:01.000 --> 00:38:04.000
something educational comes through, or whether it's things that distract

104
00:38:04.000 --> 00:38:08.000
you from learning or from developing in some way — are sensible

105
00:38:08.000 --> 00:38:12.000
images given of what a healthy, good life looks like, or do

106
00:38:12.000 --> 00:38:15.000
distorted images of what you're supposed to be come out of it instead. And now if we put

107
00:38:15.000 --> 00:38:23.000
these things together — that such a huge chunk of people's waking hours has been taken, essentially stolen, for that purpose — then it really needs to be going toward something very

108
00:38:23.000 --> 00:38:26.000
positive. And if it isn't, then we really need to be

109
00:38:26.000 --> 00:38:31.000
really careful about whether we actually want this to be happening. But what

110
00:38:31.000 --> 00:38:34.000
do you think about the argument people make in defense of this?

111
00:38:34.000 --> 00:38:58.000
That this ban is an easy way to sort of solve, or try to fix, this problem — but it doesn't actually answer the real question, which is the lack of a comprehensive federal-level data privacy law. Every state probably has some data-privacy rules of its own, some murky, GDPR-like

112
00:38:58.000 --> 00:39:22.000
setup that all differ from each other, but nothing covers the US as a whole. Which also means that even if TikTok were banned, nothing in principle stops ByteDance from simply buying that data on the open market, from third parties whose business is selling that data. So if they wanted the data that other social media collect — Instagram, Facebook, all of them —

113
00:39:22.000 --> 00:39:52.000
the American founders' companies, they could, in principle, just buy it. Although these days, since everyone's locked down their APIs and effectively blocked scraping because of AI training — once it became clear OpenAI and everyone else had scraped the entire internet's data to build models out of it — there aren't many companies left with valuable data sitting openly on the web. Places like Reddit and others are all locked down now, and that data

114
00:39:52.000 --> 00:39:56.000
gets sold to these companies separately, through a commercial deal instead.

115
00:39:56.000 --> 00:40:26.000
So in that sense, I think that kind of large-scale data-collection operation just isn't realistically possible anymore the way it was through a direct app relationship — that's a much, much more efficient, better way to collect data than trying to get it through other channels. That's actually, by the way — I don't know if people in traditional media really understand this transition, from broadcast-based media to an interest-based internet, and

116
00:40:26.000 --> 00:40:53.000
people still very much think in terms of links — that if I share my great article link here on, say, Hesari — Helsingin Sanomat, Finland's biggest newspaper — a huge amount of traffic will come in. And, in a way, the death of those link bridges has gone largely unnoticed by traditional broadcast media. People don't realize how weak these traditional media outlets actually are as platforms — even Yle, which gets 595 million euros in funding, which is

117
00:40:53.000 --> 00:41:21.000
a staggering amount — and it's still just one modest little portal that isn't really all that popular, even though you, Mikko, have been on there several times. It's not very popular among younger generations especially. So these all end up as these separate islands, and you need some kind of calling card. Maybe the modern-day equivalent of a link is a 20-second captioned short video — for this episode too, I'll be making those short clips, which

118
00:41:21.000 --> 00:41:49.000
replace the link, because a link doesn't really take you anywhere useful anymore. It's this vague, interest-based internet where bridges no longer get built — instead, content just gets thrown at everyone and floats past your eyes. Now, about that — AI is going to change this, in my opinion. That's my thesis. How? Well, now that GPT-4o uses multimodality, it's starting to become

119
00:41:49.000 --> 00:42:15.000
potentially addictive enough that you could build a really charming avatar that talks straight into your curated bubble, just as addictively. Which brings us to the point where we can actually get into Jonathan Haidt's solutions — starting from the idea that things were better before: toughen kids up by having them play in parks, take their phones away until they're 16, lock those phones in a cabinet, and set up verified identity with bank IDs. So my

120
00:42:15.000 --> 00:42:44.000
view — not a dystopian one, but an optimistic one — is that the AI interface will end up so much more addictive than short-form video that it actually ends up saving young people, and me, and you, and everyone. Well, that would require... first of all, the human mind is just built to seek out this kind of entertainment and time-wasting pretty easily — we gravitate toward content that

121
00:42:44.000 --> 00:43:11.000
is funny, entertaining, or somehow intriguing. If we could steer it, through these kinds of avatars, in an educational, useful, helpful direction, that would be really positive. I'm not yet sure whether the very best and most popular avatars will end up being like that, or whether they'll be whatever currently sells best — so for example, in the AI-companion

122
00:43:11.000 --> 00:43:39.000
sector, the biggest market right now is this kind of conversational porn — the whole point of these models is to remove the guardrails current models have around sexual topics in chat, which then makes it possible to monetize an AI girlfriend, or whatever other purpose. Right now, that's the best-selling, most profitable model out there. So, my blackpill and my whitepill — meaning my cynical view and my optimistic view — are a bit

123
00:43:39.000 --> 00:44:07.000
similar, actually — the worst case is a world where we isolate ourselves even further, where all our interaction happens with some kind of large language model, and we cut ourselves off from other people and just get even more hooked. The whitepill I heard somewhere, the optimistic view, is that hopefully at least some people would start using AI in a better way, and it could genuinely go that way, depending on the interface, of course. But for example, social

124
00:44:07.000 --> 00:44:36.000
situations, for a lot of people who aren't used to them, can be really difficult, and you can't do dry runs of them in real life. So if you haven't been toughened up to it as a kid — exactly — if you haven't really built up that resilience as a young person, learning it later can be harder, especially around people who've already picked up all those subtle ways of being, living, feeling, and expressing themselves. But if you need to take a step back, if you need to practice specifically, then why not use

125
00:44:36.000 --> 00:45:03.000
some invisible AI — visible, audible, and tangible only to you — as a sparring partner in these situations, to enhance your real-life experience and build those skills. That's the dream I'm hoping AI could move toward. What if it worked like this: someone says something to you, and then you get a moment to pause, and your assistant tells you what to say back, so that

126
00:45:03.000 --> 00:45:31.000
things go smoothly. I don't know, it's — assistive intelligence is a pretty legitimate training wheel to use. Maybe the interaction happens afterward instead, but it's basically like Terminator, just with better real-time jokes, thinking things through and telling you, 'now you should ask this.' It'd have analyzed your back-and-forth and tell you what to say next. I was a young entrepreneur back in the early 2000s — it's been a while — and there was this American mentor of mine, Bill Paulin,

127
00:45:31.000 --> 00:45:57.000
you might even remember him from back then. But anyway, with him — that's how I learned to do business in the States. We'd go to meetings, and he acted as a kind of cultural interpreter for me, meaning we always did a prep session before each meeting, going through how people react in the US, in that situation, and so on. Then we'd do the meeting, and we'd agreed on a certain signal to show when things were heading in the wrong direction.

128
00:45:57.000 --> 00:46:23.000
So Bill might scratch his ear, or tap the table, something small that nobody else noticed, but I'd know it had gone in a bad direction. Then I'd hand the floor to Bill, he'd take over and steer things back, and afterward, always after the meeting, we'd sit down and go through what went well and what didn't. I learned very quickly how to bridge into that culture from Finnish culture, and I was thinking an assistant like that could be

129
00:46:23.000 --> 00:46:52.000
really useful for bridging that kind of cultural gap — I think AI could handle that quite efficiently. Yes, indeed, yes. Great — let's wrap up with Jonathan Haidt's book recommendations, then. From what I heard: no phones for under-16s, and you get access using bank ID credentials — so it sounded like that particular battle is already lost. But the idea of exposure — kicking

130
00:46:52.000 --> 00:47:21.000
your kids out to play in the real world and take some knocks like we did — that made sense to me. In my opinion, banning phones during the school day is a really good idea, not taking them away entirely, but putting them away specifically during learning time and social time with other kids — so during breaks you're playing with each other, and in class you stay focused without those distractions during the lesson. Then your learning outcomes go back to being good, which right now are significantly weakened in schools everywhere, precisely because of this

131
00:47:21.000 --> 00:47:48.000
phone distraction. And at the same time, social skills get to develop during the school day too, through playing and roughhousing and whatnot — the point is, when you spend time with other young people, these genuinely useful human skills emerge. And then once you're home, if you want to spend the rest of the evening scrolling TikTok, fine, but at least you got that school day's worth of a break from it. Hmm, that's an interesting question, I

132
00:47:48.000 --> 00:48:17.000
agree with that, that would surely be a good thing. The question is probably how you actually get there — through legislation, or through some kind of collective agreement. Having talked with education directors of large cities on this topic, projects like this are already being launched — and honestly, all the elite schools have already moved in that direction, meaning the schools that want the best results do it anyway, because they know it has

133
00:48:17.000 --> 00:48:45.000
a direct impact on learning outcomes. That's a good point — a phone ban during lessons is much easier to achieve than one covering the whole school day, but I think it would be really important to include breaks too, so we can actually foster the growth of social skills among young people. Yes, that's probably a big deal too — Haidt talks a lot about that, about a playful childhood and specifically the value of risk-taking for young people. I don't know if this relates at all, but

134
00:48:45.000 --> 00:49:13.000
this feels like a fun podcast observation worth sharing. I was hosting an event last week, and there was a chief physician on one of the discussion panels, and I just asked a basic small-talk question about the gig — 'are you a bit nervous?' We got talking about it, like, 'well, not really, but in a positive way, a bit nervous' — that feeling before going on stage. And then he made this observation: 'I've actually noticed that a lot of young people nowadays use the word "anxious" when they talk about nervousness, or those kinds of

135
00:49:13.000 --> 00:49:40.000
situations where I'd use the word "nervousness," they use "anxious" instead.' Someone else made a comment, or I sort of nodded along, like 'what a good observation.' And then someone suggested, 'but I wonder if that's just a distinction that exists in Finnish, and maybe not in, say, US or UK English.' I mean, sorry — 'anxiety' is closer to the Finnish word 'ahdistus,' but it also slightly overlaps with 'jännitys' (nervous excitement/tension). I don't know if 'jännitys' and 'tension'

136
00:49:40.000 --> 00:50:08.000
are exactly the same thing, for instance. And then another word people use for 'jännitys' is 'excitement,' which is, in my opinion, a completely different thing. But here's the observation: fundamentally, even 'ahdistus' — this 'anxiety' feeling — might actually be a positive feeling in some sense. Sometimes it's a bit like stress; in the best case, these are the kinds of internal impulses that push you to do something, to make a change — 'I'm in a bad situation, I could do something about this' — and this bad

137
00:50:08.000 --> 00:50:36.000
feeling is what triggers it in me — so that emotional map is actually quite valid. I remember, sometimes, well, about 20 years ago NLP was such a big thing, and there we learned these techniques, sort of reframing emotions, emotional cascades — so you could be excited that you're nervous, that you're taking a risk. You can chain together positive and negative emotions, all sorts of them, turning them

138
00:50:36.000 --> 00:51:06.000
into something excitingly new, yes. Yeah, and I don't know if this is a bit of boomer-ish guesswork on my part — it might well be, coming out of my mouth — but maybe someone can back this up: are we fundamentally more inclined to always see anxiety as a bad thing? I mean, I've researched a lot into how different kinds of language emerge in these internet bubbles, and how that language is born out of a bubble's own way of communicating internally. So when people settle into different bubbles, they

139
00:51:06.000 --> 00:51:33.000
also develop a shared way of communicating there, and if you can simulate that well enough, you can pass as a member of that bubble online, of any bubble, even if you were never really in it. I've used this pretty successfully in my own online communication, pulling people from different bubbles into the same conversation by dropping in each side's own keywords and ways of talking. And because people tend to be blind to everything except

140
00:51:33.000 --> 00:51:58.000
their own way of communicating, they don't notice communication aimed at other groups the way they notice the communication that hits home for them. So in that sense, it's entirely possible that some anglicism, coming from somewhere, has taken root among young people in a certain bubble and has already, within that bubble, essentially become the word 'anxious' — so that once it's visible within that bubble, it has

141
00:51:58.000 --> 00:52:24.000
shaped how they communicate. And, well, one of our main mechanisms, in a way, is that we have this Theory of Mind, this model of the world, and you can gloss it over with emotion. I myself often say, for example, that it's never too late to have a happy childhood — you can pretty easily gild any memory with enthusiastic emotion if you just decide to, but usually people end up

142
00:52:24.000 --> 00:52:32.000
doing the opposite — through a negative, angry, or anxious culture, digging up nothing but garbage even out of good memories. Yeah, that's a good point.

143
00:52:32.000 --> 00:52:53.000
Actually, you hear this quite a lot — a pretty common counter-argument from people who feel that, collectively, we're exaggerating social media's harms nowadays, and that by focusing so hard on this, we're closing our eyes to the fact that there's a climate crisis, wars happening right now around the world, and, well, what else...

144
00:52:53.000 --> 00:53:19.000
inequality, Chinese influence through TikTok, all that sort of thing. And whoever gives that answer is, of course, revealing a lot about their own underlying biases about the world, which direction their gaze is tilted in — that these are the biggest problems right now. Of course, that's worth noting, fundamentally — but it doesn't mean the underlying mechanism is completely

145
00:53:19.000 --> 00:53:35.000
useless, or completely insignificant. I mean, the fact that we now have more information, and this somewhat distorted, but still surprisingly close-up, lens on the world's problems — a bit like what we talked about.

146
00:53:35.000 --> 00:54:03.000
At first, the problems of your own inner circle, your immediate surroundings, seem oddly distant — we protect children from those. But at the same time, all of the world's problems, all of world history's problems, the big political questions and injustices, are right there on the screen, often framed in a way that suggests you have some small part in this, or at least that

147
00:54:03.000 --> 00:54:23.000
your passivity is a sign you don't care, or something along those lines. That's a really valid point — so in practice, yes, we do need to take climate change seriously, but do we need an entire generation of young people living in climate anxiety? Did it have to become anxiety, and was there any benefit to that?

148
00:54:23.000 --> 00:54:50.000
when it comes to actually solving the problem — that's fine, if it's paired with tools for living and functioning with it. Good stuff. So, since you've watched or listened this far, go ahead and subscribe to Neuvottelija — even though Mikko keeps saying it, there's still a real benefit to it. But the biggest benefit is if you go to Google, or another search engine, and search for Isak Rautio, Mikko Alasaarela, and then Sami

149
00:54:50.000 --> 00:55:16.000
Miettinen, separately, hit enter, and only then click on these episodes. And I'd also hope that those of you who like following instructions this closely would also go watch Isak Rautio's Futucast episode, where we cover this same conversation, but probably from a completely different angle. Thanks for stopping by — and I suggest we take five more minutes on our current usage and AI, just for the Insider side

150
00:55:16.000 --> 00:55:36.000
because that's the part where followers are genuinely valuable — meaning our paying Neuvottelija subscribers — since even Google's algorithm won't tank that, because they get paid from it too. So let's jump over to the Insider side. Thank you for stopping by. Thank you, thank you.