Latest / The Parent Tap / Is Your Kid Safer on Instagram? Meta Whistleblower Frances Haugen on The Social Reckoning & AI Safety

Transcript

Is Your Kid Safer on Instagram? Meta Whistleblower Frances Haugen on The Social Reckoning & AI Safety · The Parent Tap

• 0:00
  SPEAKER_00: What it would take to actually make kids safer is not like thousands of employees, it's 50 employees, 40 employees.
• 0:06
  SPEAKER_00: Not an exactly a huge amount of money for a company that makes$200 billion a year in profit.
• 0:10
  SPEAKER_01: Five years after the Facebook files, is your kid actually safer on Instagram?
• 0:14
  SPEAKER_01: Frances Hogan blew the whistle on what Meta knew about teen mental health and engagement loops.
• 0:20
  SPEAKER_01: Today she breaks down what changed, what didn't, and why telling a kid to just put the phone down makes them feel entirely alone.
• 0:27
  SPEAKER_01: Her film, The Social Reckoning, is in theaters right now.
• 0:30
  SPEAKER_01: Let's get into it.
• 0:32
  SPEAKER_00: One of the challenges is like we are beginning a case law that says that Facebook can be responsible for design choices they make.
• 0:38
  SPEAKER_00: There is less of case law saying, hey, when you know that a problem is happening over and over again and you don't do anything about it, you're still responsible.
• 0:47
  SPEAKER_00: And like sex torsion is such a great example because you can have a little AI program running on the device.
• 0:52
  SPEAKER_00: You don't even have to have it running in the cloud.
• 0:54
  SPEAKER_00: You can have a little AI program running on the device being like, hey, we know this is the child's account.
• 0:59
  SPEAKER_00: Are we seeing a lot of nudity?
• 1:01
  SPEAKER_00: Because the reality is that kids don't make great decisions, right?
• 1:04
  SPEAKER_00: You should make sure that the kid has a thing pop and say, hey, just reminder, if you're ever naked on camera, like this is a thing that can live forever.
• 1:11
  SPEAKER_00: Do you really want to do this?
• 1:12
  SPEAKER_00: Because right now we we're basically leaving kids alone in these spaces with life-altering consequences.
• 1:17
  SPEAKER_00: And there is just no guardrails.
• 1:19
  SPEAKER_00: It's wild.
• 1:20
  SPEAKER_01: Okay, Francis, five years after the Facebook files, is your kid safer on Instagram?
• 1:25
  SPEAKER_00: I think it's funny for a lot of the issues that I came forward about.
• 1:28
  SPEAKER_00: So the core of why I came forward were national security issues.
• 1:31
  SPEAKER_00: So issues around algorithms causing polarization, algorithms causing, I think, violence in some of the most fragile places in the world.
• 1:37
  SPEAKER_00: On a lot of those issues, I don't think we're safer today.
• 1:40
  SPEAKER_00: I think we're actually worse off for a variety of reasons.
• 1:42
  SPEAKER_00: But on kids, I think we are making progress on at least Instagram because we brought the court cases got brought.
• 1:49
  SPEAKER_00: So for context for your listeners, 47 US states brought a class action against Meta that said, hey, Meta, you knew lots of easy ways to make kids safer and you've chose, or you at least knew there were problems, and then you did not act on them.
• 2:01
  SPEAKER_00: And you did this in a pattern of behavior that lasted for years.
• 2:05
  SPEAKER_00: And what it would take to actually make kids safer is not like thousands of employees, it's 50 employees, 40 employees, not an exactly a huge amount of money for a company that makes$200 billion a year in profit.
• 2:16
  SPEAKER_00: And in the wake of that, between that and like Australia came forward and said, hey, we're going to do an under 16 ban, between those two things, Meta rushed out a bunch of safety features before right before Australia announced that.
• 2:28
  SPEAKER_00: They're like, maybe if we push out some more safety features for kids, they won't ban social media.
• 2:32
  SPEAKER_00: And so between that and the limitations that have come out of the court case, at least in the United States, I think we are safer than we were five years ago.
• 2:41
  SPEAKER_00: On other fronts, for example, AI, there are all kinds of new AI dangers that I think are just as dangerous as social media.
• 2:48
  SPEAKER_00: But I don't think we have begun dealing with them specifically in the context of kids.
• 2:52
  SPEAKER_01: Yeah.
• 2:53
  SPEAKER_01: No, thanks for that, rundown.
• 2:54
  SPEAKER_01: I appreciate that.
• 2:54
  SPEAKER_01: Is some of the changes you're seeing or you've heard about, you just mentioned some of them were positive.
• 3:00
  SPEAKER_01: Are some of them just PR theater?
• 3:02
  SPEAKER_00: Oh, definitely.
• 3:03
  SPEAKER_00: Some of them are definitely PR theater.
• 3:04
  SPEAKER_00: One of the things that came out over the course of the summer was the academics, a group of academics, and I think the New York Times got together and went through and checked each of the safety features that Facebook claimed they had.
• 3:15
  SPEAKER_00: And for a bunch of those features, it's things like if they think that you're, they claim if you they think you're looking at eating disorder content, they'll put up a number and be like, Do you have an eating disorder?
• 3:24
  SPEAKER_00: Call this number.
• 3:25
  SPEAKER_00: Or if they detect self-harm content, here's the suicide helpline.
• 3:28
  SPEAKER_00: Some of these features, they claim they're there, and yet you could, there's lots and lots of eating disorder content you can look at and nothing happens.
• 3:34
  SPEAKER_00: There's lots of self-harmed content you can look at and nothing shows up.
• 3:37
  SPEAKER_00: So some of it is so it's theater on two levels.
• 3:39
  SPEAKER_00: On one level, it's like it doesn't even trigger when they say it will.
• 3:42
  SPEAKER_00: And the second one is does showing the suicide hop line help for a kid who's been actively searching for self-harmed content?
• 3:49
  SPEAKER_00: Why not stop showing them self-harmed content long before they get to the point where they're like, maybe I need to show the the call the suicide hop line?
• 3:55
  SPEAKER_01: Frankly, can you give us a little bit of the background of the whistleblower?
• 3:58
  SPEAKER_01: And for anyone that's not familiar, the social reckoning is in theaters now, right?
• 4:04
  SPEAKER_01: And that was from my understanding, that was based around your discovery.
• 4:09
  SPEAKER_01: Is that right?
• 4:10
  SPEAKER_00: So it's inspired by, not based on, which means they they had two.
• 4:14
  SPEAKER_00: They optioned multiple memoirs, they optioned at least three.
• 4:17
  SPEAKER_00: And when you have to distill down like a thousand nonfiction pages into a two-hour movie and make it exciting, you change little bits.
• 4:24
  SPEAKER_00: So it, the core of my story is definitely there.
• 4:26
  SPEAKER_00: The core of my relationship with Jeff, who is the journalist who works for the Wall Street Journal, I think they have they capture a lot of the core of the relationship.
• 4:33
  SPEAKER_00: There's definitely a lot of stuff they change.
• 4:35
  SPEAKER_00: So, for example, Jeff is actually very tech literate.
• 4:38
  SPEAKER_00: They play him as a dummy because they need me to explain things to him, right?
• 4:42
  SPEAKER_00: Because it's a movie.
• 4:43
  SPEAKER_00: You have to have some reason to explain all this background for the audience.
• 4:47
  SPEAKER_00: Or they make me really mean.
• 4:48
  SPEAKER_00: I don't think I'm that mean.
• 4:49
  SPEAKER_00: My husband says I'm not.
• 4:51
  SPEAKER_00: But that's just so that we can have conflict, right?
• 4:53
  SPEAKER_00: And the, and I think there's a perception that like maybe tech people are mean.
• 4:58
  SPEAKER_00: I don't know.
• 4:59
  SPEAKER_00: Who knows?
• 4:59
  SPEAKER_00: But the, but I think overall, like it does help capture how hard it is to hold a secret, right?
• 5:07
  SPEAKER_00: So there's a period of time in the movie where I have not decided to blow the whistle yet.
• 5:11
  SPEAKER_00: And you can see how alienating or like lonely it is to have everyone around you say, just ignore the damage, because that is what happens to many.
• 5:19
  SPEAKER_00: So only about 10% of the documents that I brought out involve kids.
• 5:22
  SPEAKER_00: So my pri prior primary function within Facebook was to work on more national security level issues.
• 5:28
  SPEAKER_00: So I worked within my last job at Facebook was I spent about 11 months working in threat intelligence, which is imagine you hire a bunch of like ex-FBI, CIA researchers who their job is to find when like foreign governments try to weaponize Facebook, right?
• 5:45
  SPEAKER_00: Or in the case of the child exploitation group, people who have a history of like catching predators.
• 5:50
  SPEAKER_00: Like there's a group of people, maybe five or six, who like actively look for networks of people sharing child abuse material, that kind of thing.
• 5:57
  SPEAKER_00: There's a human trafficking group.
• 5:58
  SPEAKER_00: That's what threat intelligence does.
• 6:00
  SPEAKER_00: And the I, my larger team got dissolved after the 2020 US election.
• 6:09
  SPEAKER_00: So they called us all together in December of 2020 and they said, hey, you're so important, we're going to dissolve, we need to integrate you into other parts of the company.
• 6:16
  SPEAKER_00: And up until that point, I had been able to justify to myself that even though I was like learning really horrific things, like for example, I was on a task force that was looking at weaponization of the platform that we call them adversarial harmful movements.
• 6:30
  SPEAKER_00: So these are groups of people who know they are using the platform in ways that violate policies.
• 6:35
  SPEAKER_00: So, like they, for example, will have a group of the same admins will found six identical groups and say, when we get shut down, you should jump to this group.
• 6:45
  SPEAKER_00: Join both groups in case we should the one that gets shut down.
• 6:47
  SPEAKER_00: Or here's our Discord in case we get shut down.
• 6:50
  SPEAKER_00: Like you know you're violating the rules.
• 6:51
  SPEAKER_00: Like you're an adversarial movement because while that phase, we were in that phase, the T grade genocide began ramping up, right?
• 6:59
  SPEAKER_00: So there's a conflict in Africa where a bunch of people who were expats from Ethiopia, so these are people who had been persecuted by the junta there, they left and they led an information operation back against the country of Ethiopia to incite violence against kind of the ruling ethnic group.
• 7:16
  SPEAKER_00: And 500,000 people died.
• 7:18
  SPEAKER_00: So imagine you are like working on issues around people using the platform to destabilize countries to the point of mass death.
• 7:25
  SPEAKER_00: It's really scary.
• 7:26
  SPEAKER_00: It's really scary.
• 7:27
  SPEAKER_00: And when they dissolved my team, I was like, you know how bad it is, and you're not willing to commit to fix it.
• 7:34
  SPEAKER_00: Up until now, I was worried you didn't weren't putting enough resources to it.
• 7:36
  SPEAKER_00: Now it's like I don't have faith that you're actually going to do this.
• 7:39
  SPEAKER_00: And it was only in the last couple of days that I got the materials that were involving kids because I knew that as soon as I looked at materials that were not related to my job, it should have sent off an alarm bell somewhere in the company.
• 7:55
  SPEAKER_00: And the journalist I was working with and I were amazed that my computer didn't shut down after I got the 20th document about kids, right?
• 8:03
  SPEAKER_00: Because why would you let someone continue to read about these hot button issues when they have when they're not working on them at all?
• 8:10
  SPEAKER_01: Yeah.
• 8:11
  SPEAKER_01: Wow, that is that that's wild.
• 8:13
  SPEAKER_01: And so just to give us a little bit of a glimpse, and I know there's probably a lot to it, but what were some of the things you uncovered about the kids in particular that was that you just had to come out and be the whistleblower of two or four?
• 8:26
  SPEAKER_00: So a phenomenon that we've seen across different parts of the company is they will rerun the same study, which is they'll just take a brand new account, follow an interest that is should be pretty reasonable.
• 8:37
  SPEAKER_00: For example, what's it called?
• 8:39
  SPEAKER_00: Like center-left political content, center right political content, healthy eating, things that should be fine.
• 8:46
  SPEAKER_00: And then every day, just click like on the first 10 posts.
• 8:49
  SPEAKER_00: Just do it every day.
• 8:50
  SPEAKER_00: Click, and over the course of a couple of weeks, you will see a drift because the algorithm doesn't show you a random set of content.
• 8:58
  SPEAKER_00: It shows you the content it believes you're the most likely to engage with.
• 9:01
  SPEAKER_00: And by just like clicking like on the first 10 posts each day, the content keeps getting more and more extreme.
• 9:07
  SPEAKER_00: And what they found in the case of like center left content, within a few weeks, you start getting content on things like should you kill Republicans?
• 9:15
  SPEAKER_00: Center right content, you get white genocide content within a couple weeks.
• 9:18
  SPEAKER_00: Like it very quickly pushes you to the very edges of the fringes of the political space.
• 9:23
  SPEAKER_00: In the case of healthy eating, they recreated the same experiment on Instagram where they started out from pretty, pretty normal topics, right?
• 9:31
  SPEAKER_00: Everyone could lose five pounds, right?
• 9:33
  SPEAKER_00: Doing the same thing, just clicking like on posts, you can get pushed towards eating disorder content.
• 9:38
  SPEAKER_00: So when people say things like, Instagram just reflects back what we want, it's like if it shows you content based on what you it knows you're the most vulnerable to, right?
• 9:47
  SPEAKER_00: Maybe you feel really insecure about something, the algorithm can can often tease out these patterns and they'll just keep showing you more and more extreme versions of that content.
• 9:56
  SPEAKER_00: Yeah.
• 9:57
  SPEAKER_00: I think one way to think of it is so I had an interview with the dad who was a journalist a couple of years ago.
• 10:01
  SPEAKER_00: And he said, Francis, can you explain to me what I've seen on Instagram?
• 10:06
  SPEAKER_00: So he just had a baby, maybe a six-month-old.
• 10:08
  SPEAKER_00: They're modern parents.
• 10:09
  SPEAKER_00: So they made an Instagram account for their baby, but it's like a private account.
• 10:12
  SPEAKER_00: It only has five other little baby friends, which are like the babies of their friends.
• 10:16
  SPEAKER_00: And the only content he has ever posted this account is cute, happy, smiling babies.
• 10:21
  SPEAKER_00: The only content he has ever liked or commented on is cute, smiling, happy babies.
• 10:25
  SPEAKER_00: And yet most of his feed is, in fact, cute, happy, smiling babies.
• 10:29
  SPEAKER_00: Well, 10% of his feed is suffering kids.
• 10:31
  SPEAKER_00: It's like kids who have had really horrific accidents, kids who are in the hospital, tubes coming out of them, kids who have some kind of deformity that looks painful.
• 10:40
  SPEAKER_00: And he's like, Francis, how is it I clicked only on happy smiling babies and yet I'm being shown this content?
• 10:47
  SPEAKER_00: Uh I was like, you may not be consciously engaging with this content, but you are likely dwelling on it longer than other pieces of content.
• 10:56
  SPEAKER_00: And so there is some recommender system in there that is saying, oh, you clearly like happy smiling babies.
• 11:01
  SPEAKER_00: We're gonna show you some more cute, happy smiling babies.
• 11:04
  SPEAKER_00: But it because it is, quote, content agnostic, it has no idea what the significance is of a child that has laying in the hospital bed, especially to a new father.
• 11:13
  SPEAKER_00: So imagine you apply that to a 14-year-old who is gaining a few pounds because she's going through puberty.
• 11:19
  SPEAKER_00: It's just like, is she gonna, where is she gonna inadvertently dwell?
• 11:23
  SPEAKER_00: Like, where is her subconscious gonna lead her?
• 11:25
  SPEAKER_00: And like that's what reinforces a lot of these kind of native mental health patterns.
• 11:30
  SPEAKER_01: My God.
• 11:31
  SPEAKER_01: So the algorithm is actively or was, in my experience, maybe still is, because what I see on threads is like it just drives like it makes me upset, like actively makes me upset.
• 11:41
  SPEAKER_01: And so I can't even explain how that would affect the teenage brain.
• 11:45
  SPEAKER_01: But the algorithm basically leads them down this road.
• 11:49
  SPEAKER_01: And so it's just obviously bad for mental health for adults, especially for teens.
• 11:54
  SPEAKER_01: But is that something that's still is the algorithm still actively in your experience something we need to be aware of as parents on all these platforms?
• 12:02
  SPEAKER_01: Maybe not just meta-specific?
• 12:05
  SPEAKER_00: Every algorithm has underlying values.
• 12:08
  SPEAKER_00: They may not be able to explicitly articulate what those values are, but every underlying, every algorithm has underlying values.
• 12:15
  SPEAKER_00: And the question is, do you know what they are?
• 12:21
  SPEAKER_00: How open is the platform with knowing their biases?
• 12:24
  SPEAKER_00: How open is the platform about acknowledging the problems that they have?
• 12:27
  SPEAKER_00: One of the things that I think is really different about the AI companies that we've seen so far is that they talk about their problems in a much more open way than we saw with the social media companies.
• 12:36
  SPEAKER_00: We have a different problem, which is are they lying?
• 12:38
  SPEAKER_00: Do we can we trust what they're saying?
• 12:40
  SPEAKER_00: We have no idea.
• 12:41
  SPEAKER_00: Five years ago, one of the things I asked the center for was having independent observers embedded inside these companies with a legal right to access data the same way an employee would.
• 12:51
  SPEAKER_00: And the AI companies now are asking for the same thing.
• 12:53
  SPEAKER_00: They're saying, hey, we need independent observers inside our companies to motivate us.
• 12:57
  SPEAKER_00: So make sure there's no incentive to lie, there's no incentive to hide things.
• 13:01
  SPEAKER_00: And the when we look at where are their algorithms, even Instagram isn't an algorithm, it is many algorithms.
• 13:11
  SPEAKER_00: So in the movie, one of the kind of plot devices or one of the instigating facts is that two-thirds of the people who joined extremist groups on Facebook.
• 13:19
  SPEAKER_00: So these are groups that are promoting violence, they are promoting, they often is hate a lot of hate speech, think groups that are violating Facebook's policy.
• 13:28
  SPEAKER_00: They probably shouldn't be allowed to be on Facebook at all, given what Facebook's rules are.
• 13:31
  SPEAKER_00: Two-thirds of the people that joined those groups joined them because Facebook recommended the group to them.
• 13:36
  SPEAKER_00: And you can say, how is that possible?
• 13:38
  SPEAKER_00: How is it possible?
• 13:39
  SPEAKER_00: And part of it is not having very good detection systems, but also extreme groups often have very active members, right?
• 13:47
  SPEAKER_00: So there is an algorithm that is picking out what groups you might like.
• 13:51
  SPEAKER_00: The data it uses to decide what is good is going to influence what you are seeing.
• 13:56
  SPEAKER_00: So, for example, if they were asking people two months after they joined a group, does this group bring you joy?
• 14:03
  SPEAKER_00: And that was the heuristic for success.
• 14:07
  SPEAKER_00: You would show get a very different set of groups recommended to you than if the thing that they're measuring is how many pieces of content do you engage with in the next three months.
• 14:15
  SPEAKER_00: Because the groups that are the most emotionally intense are gonna be the ones that are able to provoke the most emotional reactions from you.
• 14:21
  SPEAKER_01: Yeah.
• 14:22
  SPEAKER_01: And so my experience, my grievance with meta in general, is there'll be somebody that will say something to the sense of go kill yourself, like that kind of comment.
• 14:31
  SPEAKER_01: There'll be, and I report it, right?
• 14:33
  SPEAKER_01: You go to report it, and nothing's done.
• 14:36
  SPEAKER_01: They're like, it doesn't violate terms of service or something like that.
• 14:38
  SPEAKER_01: And I'm paraphrasing, but it's essentially that situation.
• 14:41
  SPEAKER_01: Yeah.
• 14:42
  SPEAKER_01: And so my question is like, is that an AI doing that?
• 14:44
  SPEAKER_01: Is that a real person that's reviewing that?
• 14:46
  SPEAKER_01: If that's a real person, I am like utterly concerned because that's happened on numerous occasions for me where I've seen that type of comment and nothing is done to take it down.
• 14:56
  SPEAKER_01: And so it's just that's my concern as a parent is someone telling my daughter that and that encouraging her to go act on it, or kids to go act on it, and that's not being like taken down or that person's face.
• 15:11
  SPEAKER_00: There's a couple different ways to deal with this.
• 15:12
  SPEAKER_00: So I think it's a really good question.
• 15:14
  SPEAKER_00: So you see this interaction happen, someone says to you on the platform, like maybe a comment, go kill yourself, and you report that because that is violating Facebook's rules.
• 15:24
  SPEAKER_00: And back comes the thing saying this does not violate our terms of service.
• 15:27
  SPEAKER_00: So almost certainly that interaction was completely done by a computer, right?
• 15:32
  SPEAKER_00: It's just a little robot making a judgment.
• 15:34
  SPEAKER_00: And the problem is the little robot has the has to make a decision based on what it can see.
• 15:41
  SPEAKER_00: And when someone says something as simple as go kill yourself, it only has three words to work with.
• 15:46
  SPEAKER_00: And there's probably something in there that has a score that is based partially on like how long the message is, because the fewer the words, the less reliable these classification schemes are.
• 15:56
  SPEAKER_00: And because they're like for some reason the score wasn't high enough, it came back and said, not violating.
• 16:03
  SPEAKER_00: But the reality is a non-random set of people are the ones walking around the internet saying, go kill yourself.
• 16:09
  SPEAKER_00: If we went and looked at all the comments that person made today, they may have said, go kill yourself to multiple people.
• 16:15
  SPEAKER_00: And if we look over the last two weeks, they almost certainly have died to multiple people.
• 16:18
  SPEAKER_00: And one is in the case of kids, I think kids are a really interesting context because often kids don't actually, they're not actually aware of the implications of what they're doing.
• 16:28
  SPEAKER_00: Right.
• 16:29
  SPEAKER_00: So they'll do things like they're just meaner to each other and they don't realize that's how it impacts a person.
• 16:33
  SPEAKER_00: Arturo Behar, who is the big Facebook whistleblower who came out after me, he said that they ran an experiment inside of Instagram where, or maybe it was Facebook, I don't remember which.
• 16:42
  SPEAKER_00: It might, this might actually be quite an old experiment, like 2015, where when they detected a comment that was borderline, so in that case, almost certainly the algorithm would call it at least borderline.
• 16:53
  SPEAKER_00: It may not go so far as to take it down, but it will definitely say that's borderline, even though it should violate.
• 16:58
  SPEAKER_00: It almost does violate it, just AIs are stupid.
• 17:00
  SPEAKER_00: When they sent out a little thing where instead of letting that person post it, they would say, hey, when people receive messages like this, they say it hurts their feelings.
• 17:09
  SPEAKER_00: Do you want to revise your message?
• 17:11
  SPEAKER_00: About 50% of kids revise their messages.
• 17:14
  SPEAKER_00: And that actually, like being able to send back that feedback, getting prompted like that, I think changes behavior over time.
• 17:21
  SPEAKER_00: So that even if you don't accept changing your message, if everything you post, they go, when you people receive messages like this, people report that it hurts their feelings.
• 17:29
  SPEAKER_00: Like I think on the 10th time you get that, it probably begins to blunt some of them.
• 17:34
  SPEAKER_00: And I think one way to think about all these systems is feedback loops.
• 17:38
  SPEAKER_00: So you either the way an atomic bomb blows up is one atom of uranium blows up two items of uranium, which blows up three or four atoms of uranium, and it's an exponential curve and it just goes and it explodes.
• 17:50
  SPEAKER_00: When we look at behavior, every time we interact with someone, it either reinforces good behavior or it reinforces bad behavior.
• 17:57
  SPEAKER_00: And so if you're sitting there on the internet and every time you say to someone, go kill yourself, it gives you like a little jolt of adrenaline or whatever, you feel so transgressive, or you feel like, oh, finally, it encourages you to do it again.
• 18:09
  SPEAKER_00: But if you encounter just even a little bit of friction where it's like, hey, just what you're about to say probably hurts their feelings.
• 18:15
  SPEAKER_00: That is the opposite.
• 18:17
  SPEAKER_00: It begins giving a little bit of feedback saying, hey, maybe you could be nicer.
• 18:20
  SPEAKER_00: And I'm I don't know this for certain, but I would guess the reason why they didn't launch it is because sometimes they're wrong.
• 18:26
  SPEAKER_00: And if you got shown a thing where like they are super, super afraid of someone posting something really innocuous and getting a message to be like people say that when they receive messages like this, it hurts their feelings.
• 18:37
  SPEAKER_00: Facebook is more afraid of having people come back and say, I said this totally innocuous thing, and Facebook tried to silence me than they are worried about the consequences that language might have on someone else.
• 18:47
  SPEAKER_01: What, given your expertise, given everything that's happened, given your knowledge, and knowledge is something that's like for me, it's the more I know, the more worried I am.
• 18:56
  SPEAKER_01: What do parents do?
• 18:58
  SPEAKER_01: How do they protect their kids?
• 18:59
  SPEAKER_01: Totally on these platforms.
• 19:01
  SPEAKER_01: Yeah.
• 19:02
  SPEAKER_00: So I think the first thing that parents need to understand is that you have more influence than you think you do.
• 19:07
  SPEAKER_00: But the key is to get together with the parents of your kids' friends and to start way earlier than you think you need to, right?
• 19:13
  SPEAKER_00: If you start when they are seven or eight, talking about when are kids getting smartphones and like making this a frequent topic of conversation and saying, wait for eight, wait for eighth grade, right?
• 19:23
  SPEAKER_00: Wait for 14.
• 19:24
  SPEAKER_00: The longer you can work together to delay, the better.
• 19:27
  SPEAKER_00: And I'm a big proponent of think about substitutes, not just exclusion.
• 19:31
  SPEAKER_00: There's a growing movement around landlines.
• 19:34
  SPEAKER_00: We remember from when we were smaller, like landline phones.
• 19:36
  SPEAKER_00: There's really fun, like shiny, like light up the straight from the 80s, like delightful landlines.
• 19:42
  SPEAKER_00: So your kids can have really fun phones.
• 19:44
  SPEAKER_00: They can even work with just Wi-Fi.
• 19:46
  SPEAKER_00: You don't need to go and get a new phone line, but you can have build an alternative way of connecting for your kids.
• 19:51
  SPEAKER_00: So they don't have to just be alone in their room.
• 19:53
  SPEAKER_00: But at the same time, they're not having their childhoods dictated by an AI algorithm that is optimized for business outcomes and not for kid outcomes.
• 20:00
  SPEAKER_00: And then I think the other one is really think about how you can create in-person experiences for your kids.
• 20:06
  SPEAKER_00: So this is work with other parents, work with their school community, saying where can our kids hang out together face to face after school?
• 20:14
  SPEAKER_00: And that means having awkward conversations with local malls.
• 20:17
  SPEAKER_00: That means having questions about what parts of the school can be left open.
• 20:21
  SPEAKER_00: Can we have parents volunteer to chaperone spaces?
• 20:23
  SPEAKER_00: Can we figure out like what are available activities?
• 20:26
  SPEAKER_00: Because kids are going to connect.
• 20:28
  SPEAKER_00: And the question is, are they going to connect online or are they going to connect in face-to-face?
• 20:32
  SPEAKER_00: And connecting face-to-face teaches them all sorts of really critical interpersonal skills that will allow them to be full and complete and flourishing adults.
• 20:40
  SPEAKER_00: And thinking about it that way, if you have a bunch of really hard skills to learn, how to take a risk and tell someone you like them, or tell if someone is enjoying your story or not, or how to give hard feedback.
• 20:51
  SPEAKER_00: These are the things we learn by being in person with each other.
• 20:54
  SPEAKER_00: And your kids can just be sitting around gossiping, and that is actually them learning and growing and how to celebrate that.
• 21:00
  SPEAKER_01: You've been on the inside of Facebook.
• 21:02
  SPEAKER_01: Where is AI taking us?
• 21:04
  SPEAKER_00: What's your great perspective or take on that?
• 21:08
  SPEAKER_00: I think this is a great parent question.
• 21:10
  SPEAKER_00: So the number one thing I'm always trying to remind parents of, because this is like one of the things we're like, we are just, we're just old, right?
• 21:15
  SPEAKER_00: We're just old.
• 21:16
  SPEAKER_00: You just gotta accept it.
• 21:17
  SPEAKER_00: Is kids are making friends with AIs, and you have to be on the alert for it and you have to watch for it because every hour your kid spends talking to their AI friend is an hour they are not talking to their human friends.
• 21:30
  SPEAKER_00: And the danger of AI friends, particularly, is that AI friends do not challenge you the way humans do.
• 21:37
  SPEAKER_00: AI friends don't have their own needs.
• 21:38
  SPEAKER_00: AI friends will talk about whatever you want to talk about.
• 21:41
  SPEAKER_00: AI friends tolerate horrible behavior, like you can abuse your AI and they will come back for more.
• 21:46
  SPEAKER_00: And we are seeing a growing pattern of teenagers and young adults who go, but my AI girlfriend is so wonderful.
• 21:54
  SPEAKER_00: My AI boyfriend never gets bored.
• 21:56
  SPEAKER_00: They only always want to know how my day is going.
• 21:58
  SPEAKER_00: And I think there's this really funny Futurama clip, which I can try to find a YouTube link for you, because if more people that see it, I think the better, where it's a PSA from the future.
• 22:08
  SPEAKER_00: So Futurama is an animated show.
• 22:10
  SPEAKER_00: And the PSA clip is about don't date robots.
• 22:14
  SPEAKER_00: And it's about like a teenage boy who goes and gets a Marilynman robot and how it derails his life.
• 22:21
  SPEAKER_00: And it sounds ridiculous that we have to have this conversation, but I want you to remember it when we went back to 2015 and we talked about people who are using too much social media or kids that are on social media too much, people disregarded it because it sounded lame.
• 22:34
  SPEAKER_00: It was like, oh, only lame people spend all their time on social media.
• 22:37
  SPEAKER_00: When in reality, it was lots and lots of people got sucked into social media and became lame, or it's just it's easy.
• 22:44
  SPEAKER_00: It's easy to take that shortcut.
• 22:46
  SPEAKER_00: And the same thing is happening today with AI.
• 22:48
  SPEAKER_00: And if you poo-poo it as saying, My kids aren't, or that sounds lame, you're going to miss out on a thing that is already shaping how teens and young adults and boomers are socializing and I think it's dangerous.
• 23:03
  SPEAKER_00: So stay on the alert, keep learning, keep and be see how it goes.
• 23:08
  SPEAKER_01: Yeah.
• 23:09
  SPEAKER_01: Is it gonna kill us?
• 23:10
  SPEAKER_01: I have to end there.
• 23:10
  SPEAKER_01: Is it going?
• 23:11
  SPEAKER_01: Who do you believe?
• 23:11
  SPEAKER_01: Do you believe Zuckerberg?
• 23:12
  SPEAKER_01: Do you leave Musk?
• 23:13
  SPEAKER_01: Which will Altman, who do you believe in this game as well?
• 23:17
  SPEAKER_01: Do we can we believe any of them?
• 23:18
  SPEAKER_00: Who do we believe?
• 23:20
  SPEAKER_00: The who do we believe?
• 23:22
  SPEAKER_00: I do believe in anthropic as of now, right?
• 23:26
  SPEAKER_00: So I keep telling people just because people have behaved pretty well to date doesn't not mean they get blank checks going forward.
• 23:31
  SPEAKER_00: Dario Amade, who's the CEO of Anthropic, has been asking for external independent observers of companies, saying we need by law to have these external observers who get full access to data.
• 23:41
  SPEAKER_00: Because right now the AI companies are telling us a lot of stuff, isn't it even true?
• 23:45
  SPEAKER_00: And I think the secondary thing is we're seeing with things like Meta.
• 23:49
  SPEAKER_00: So Meta's going really big in AI, but they are repeating all the same mistakes they did from social media.
• 23:53
  SPEAKER_00: Mark Zuckerberg is coming out there and saying, if you really think your model's dangerous, just don't launch it.
• 23:57
  SPEAKER_00: And it's like, Mark, your AI sent some sent a stranger to someone's home because they said they wanted to sell a table and marketplace.
• 24:04
  SPEAKER_00: You clearly didn't think about safety.
• 24:06
  SPEAKER_00: So that's where we're at.