YT YouTube2Text

YouTube2Text → Transcripts → Facebook Whistleblower Frances Haugen On Product, Competition, And Transparency

Facebook Whistleblower Frances Haugen On Product, Competition, And Transparency — Transcript

by Alex Kantrowitz · 12,505 words · 1,852 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00all right everyone welcome to the big
0:02technology podcast a show for
0:04cool-headed nuanced conversation of the
0:06tech world and Beyond we are live here
0:08at unfinished live in New York City and
0:10just to show you that there's actually a
0:12live audience before us I'm going to ask
0:15them to let us hear it you guys got to
0:17be loud because these microphones won't
0:18pick it up unless you give it your all
0:20let's hear you guys
0:25all right we got to keep that energy up
0:28through this whole thing all right I
0:29don't want to see anyone walking out you
0:31don't always get a laugh track so you
0:32know you've got to lean in while you
0:33have one that was good we're gonna we'll
0:35put that uh Ed any good moment along the
0:38podcast we'll put that now everybody boo
0:39no I'm kidding we're here with Francis
0:42Haugen uh you might know her as the
0:44person behind the Facebook files the
0:46Facebook papers
0:47it's not quite know which which name to
0:49use but they're the series of
0:50Revelations
0:51um that showed a lot of things about
0:53Facebook's business
0:54that really were not great and
0:58um those Revelations happened a year ago
1:01and they've sparked a series of
1:02discussions and
1:04um much more informed conversation about
1:06the platform than I think we had
1:09previously as a journalist I'm really
1:10grateful uh to have been able to comb
1:12through some of the material that you
1:14put out there because it finally allowed
1:16us to put some some data and some actual
1:19internal research Behind these things
1:20that people were trying to you know
1:23shoot from the hip and say were
1:24happening and we didn't really know
1:25whether that was true or not so first of
1:27all thank you for revealing all that
1:29stuff
1:30second let's just take a look at at
1:32meta's uh stock price right now
1:35um I took a look about September 13th is
1:38when the beginning of the Facebook file
1:39series of Articles came out in the Wall
1:42Street Journal
1:43um since that week meta stock it's now
1:45Meta Meta stock is down 59 wow so uh I
1:51only check it occasionally so I have one
1:54Twitter follower who messaged me
1:55messages me every time it goes up and
1:57he's like oh no the stock is recovering
1:58that I look at it again I'm like oh it's
2:00actually down that yeah that's called
2:02unnecessary paranoia yeah
2:05um so I guess looking at the way the
2:08stock has performed
2:10um mission accomplished
2:12um you know sometimes people ask me
2:14questions like uh you know do you think
2:17you won and and I feel like that really
2:20um misses a large part of the criticism
2:23I've given which is um I feel like I'm
2:25fairly characteristic for whistleblowers
2:27in that when you when you look at it
2:29statistically overwhelmingly
2:30whistleblowers are conscientious people
2:32who actually care a lot about their
2:34companies and I really believe that uh
2:37Facebook has problems that the only way
2:40it can successfully solve them is by
2:42working on them collaboratively and so
2:44my intention was never to have the stock
2:46price go down my my My Hope was that
2:49Facebook could get the support it needs
2:51to be long-term successful and and
2:54unfortunately the stock has gone down
2:55because I think the public has looked at
2:57how they responded like the biggest
3:00thing they did fresh out of the gate was
3:02they're like actually we're in a
3:03metaverse company now and in general I
3:05think investors get quite scared when
3:07you say you have a giant problem and
3:08they're like oh shiny
3:10you know it's not not really a great uh
3:12forward-looking indicator and this is
3:15also coincided it was the top of of the
3:17tech uh market cap era it's crashed
3:20we've had there are other factors
3:22involved we also have had the FED raise
3:24rates in large Fallen substantially more
3:26than the other social media companies
3:27right do you attribute that to your
3:29Revelations
3:31um so in our SEC complaint we said um
3:34Facebook's evaluation so I actually want
3:36to look back at five years of Facebook
3:37stock price so I'm a data scientist and
3:39so I I view the world through data and I
3:41looked at every time I want to actually
3:43went and Google like specifically to
3:45Target Googling for every time Facebook
3:47stock price declined more than five
3:49percent versus the NASDAQ over a
3:52five-year period before I came out or
3:54before the first article came out
3:55and there's a very limited number of
3:57times that happen and there's about 27
4:00um and uh overwhelmingly the things that
4:03happened when the stock price declined
4:05more than the NASDAQ declined
4:07um were Facebook had to decline users
4:09Facebook had some kind of security
4:12incident where they had to come out and
4:13say we're gonna have to spend more on
4:15safety
4:16um or it was implied they were gonna
4:17have to spend more in safety
4:18um and it was something like on the
4:20order of 65 of the time when you had
4:22those deviations from the NASDAQ
4:25um our SEC complaint has said the price
4:28was overvaluated was over inflated
4:30because users use the product who would
4:33not have used it had they known the
4:34truth advertisers advertised would not
4:37have advertised had they known the truth
4:38and Facebook underspent on safety more
4:42than they would have than if they had
4:44been transparent and since then we've
4:46seen it to client users we've seen on
4:48them having to increase the amount they
4:50spent on safety by two to three times
4:51which is huge compared to the Past times
4:53they've increased safety and so we can
4:55and we'll never be able to know exactly
4:57what fraction of this is because of the
5:00transparency but I think there's a good
5:02chance that the the information did
5:04cause the stock to correct back to
5:07something closer to its natural
5:08evaluation a skeptic would say that that
5:11Apple's anti-tracking moves probably had
5:13a larger impact what do you think about
5:16that I think unquestionably they had an
5:18impact as well I I
5:19never want to assess
5:21relative contribution because as
5:24a quantitative people are naturally
5:26um
5:28I always try to be very precise in what
5:30I say and I have no data to indicate
5:31what the relative contribution is okay
5:34it'll be interesting I know you can't
5:35speak about the SEC uh or discussions
5:40um Kara Swisher tried that a code didn't
5:41get anywhere so
5:43I won't try to press that I actually
5:44just don't like that yeah like they've
5:45talked to me once okay and apparently
5:48that's fairly standard right
5:50um they I usually try to not talk to
5:53Witnesses too many times because if you
5:55ever don't say exactly the same thing
5:57that can actually make it harder for
5:58them
5:59um on and they usually go dark for long
6:03periods of time while they're building
6:04their case and so I I genuinely don't
6:06know much about what's going on okay
6:09so as Facebook struggles
6:12um whether it's because his stock price
6:13has declined due to these Revelations or
6:16whether apple is kneecapping it or maybe
6:18Apple's having more success because of
6:20the revelations that you brought out
6:22there's this other there's a counter
6:24current to that and that's that
6:26competitors get stronger
6:28the one that really seems to be doing
6:30quite well right now is Tick Tock and
6:33you know a while ago people would say
6:35okay well
6:36um the fact that Chinese companies could
6:39take the place of Facebook is just a big
6:41Tech talking point that they're trying
6:43to you know divert attention to their
6:45own actions but now it really is
6:46happening
6:47and so what have you what do you make of
6:49the fact you know is there any Credence
6:51to the fact that as a company like
6:53Facebook gets weaker a company like Tic
6:55Tac can emerge and actually leave us in
6:57a worse position than we were previously
7:00so I think tick tock's emergence is a
7:02great example so throughout you know I
7:05think basically all of capitalist
7:06history we have seen time and time again
7:08that if consumers get to choose between
7:09products that make them happy and
7:11products that make them sad
7:13as a general rule people choose products
7:14that make them happy over products that
7:16make them sad
7:17and um I when I look at the things that
7:20make Tick Tock sticky you know I think
7:22there are are genuine questions we need
7:24to ask right so part of the reason why
7:26Tick Tock makes you happy is Tick Tock
7:28emerge the the fact that the only
7:31successful competitor to Facebook
7:33emerged from the only place Facebook
7:35wasn't allowed to play
7:38right like Facebook went into all the
7:40other corners of the world and sometimes
7:42even bribed the locals to say let us
7:44become your internet right they paid for
7:46people's data to get people to converge
7:48onto it
7:49um the fact that it emerged from China
7:51meant also the only competitor that
7:53could emerge was one that met Chinese
7:56goals
7:57and China created a social media
8:00platform that focused on Dancing on
8:01comedy on lighthearted things because
8:04those things are often not the same as
8:05politics
8:07right
8:08um and at the same time they've created
8:11a product that makes people feel good
8:13and so if you have a situation like that
8:15someone was eventually going to come
8:17along with a product or even be real
8:19feel real makes people feel better than
8:21Facebook does
8:23um and so I I think that's the the real
8:24thing you looked at a lot of Facebook
8:26research some of that research said uh
8:28that people feel worse if they scroll
8:31and don't participate just passive
8:33consumption makes people feel worse
8:35if that was the internal Facebook
8:36research how do you think it is that
8:38Tick Tock a app that literally has
8:41people scroll passively uh that is
8:43making people feel good but they but
8:45they actually don't like like you you
8:47have to you have to choose to move on
8:49with Tick Tock if you if you are truly
8:51passive like you're not engaging at all
8:52if you're not choosing to do a next
8:54gesture it just replays the video right
8:56and and I think it's also a thing of um
8:59a tick tocks algorithms because it is
9:01like I think I keep trying to flag for
9:03people this Tick Tock is a very
9:04manipulative experience algorithmically
9:06right they bias towards happiness they
9:09bias towards humor like they don't want
9:12you talking about sad topics because
9:14political topics often are sad topics
9:16right and um and so you know funny
9:19dances
9:21if you can be passively you can be not
9:23really engaging not having a
9:24conversation and watching funny things
9:26and you're going to feel laughter
9:28laughter makes you feel better is that
9:31what your Tick Tock feed is mostly funny
9:32dances you know I get a lot of um home
9:35restoration
9:37um
9:40really like cake decor I don't know why
9:42I can't have gluten who knows
9:45um they really think I really like cake
9:48decorating and they really think I like
9:49renovating homes so that's like a large
9:51fraction of my feed I don't know what
9:52this says about me but mine is all like
9:55couples finding their partner cheating
9:57really yeah they are great videos you
10:00know you know uh you could you could
10:02thumbs down some stuff and I like them
10:04oh well you know
10:06so well go with God my friend indeed I
10:10will
10:11um so let's get back to that question
10:12though about whether we're worse off
10:14with with Tick Tock Dominic I mean yeah
10:16okay it's making us happier the other
10:18side of it is feel happier yeah I don't
10:21know if it like existentially that's
10:22true yeah we could we could go back I
10:24think sucking our time in we'd probably
10:26be happier gardening but you know that's
10:28true yeah
10:29um so so but speaking about that
10:32um okay let's say feel happier
10:34um but it's doing keystroke logging or
10:37it's at least alleged to do that in its
10:40in-app browser yeah and which is a
10:42pretty big deal and also we don't fully
10:45know the extent of the oversight that
10:48the Chinese government may or may not
10:49have yeah over that app so if American I
10:53know we're going to get into Global
10:54stuff well let's talk about U.S if
10:56American social media users
10:59you know what let's just go Global if
11:01global social media users are in an app
11:04that has these major liabilities
11:06isn't that worse than than being inside
11:09Facebook inside Facebook well I think
11:11it's this question of um you know like
11:13what we I think we need to have a
11:19I I think we should have Facebook of
11:212008 like if what I was gonna if we're
11:23going to do any kind of magical
11:24hypotheticals you know social media that
11:26is about our family our friends our
11:28communities contact connecting with
11:30people is better than anything where you
11:32have hyper virality
11:34um the things that are are good about
11:35Tick Tock are Tick Tock acknowledges
11:38they have power
11:39and Facebook really doesn't want to
11:41acknowledge they have power right they
11:43like to be like oh no we can't do
11:44anything like we have no ability to we
11:45have no ability to do anything
11:47um the secondary thing is
11:49um you know if we had to choose between
11:52you know an app that has had two
11:54genocides so you know we have 200 000
11:56people dead in Myanmar this is not my
11:58opinion the UN has a multi hundred page
12:00report detailing what Facebook did those
12:02negligent if we see the same thing
12:04happen again in Ethiopia
12:07um with lots of warning if we see this
12:09happen over again I think there's this
12:10question of at some point we have to
12:12decide
12:14when do we have to hold Facebook
12:15responsible for being an app that seems
12:17to have a pattern of causing genocides
12:19because I have not seen any Tech talk
12:21genocides yet
12:24um
12:26there have been genocides before
12:28Facebook it's
12:29right so do we do we now in the age of
12:32Facebook where do we now put the blame
12:35for all genocides on Facebook or no I
12:37think there's I think there's ways of
12:38acting responsibly in a space so for
12:41example
12:42you know uh in 2002 excuse me 20 2020
12:462020 2020. in April 2020. in preparation
12:50for the U.S election uh they held an
12:52internal working group it's like 60
12:54people I was one of the people on it
12:56um who were tasked with coming up with
12:58quote soft interventions to prepare for
13:00the election
13:01right so a soft intervention is
13:02something like uh you um instead of
13:06taking down account taking out a post
13:07you say hey maybe you should pop up a
13:11thing saying hey you know you didn't
13:13click on a link you didn't actually
13:14click through on this link you want to
13:15click on it first because like putting a
13:17little moment of intentionality before
13:19someone re-shares something actually
13:21substantially decreases misinformation
13:22you haven't censored anyone it's like a
13:25soft intervention and one of the things
13:26that was flagged was
13:28um there was a single parameter in
13:31meaningful social interactions so in
13:33back in 2018 Facebook switched from just
13:35seeing how long they could keep you on
13:36Facebook to seeing could they drive you
13:39to do interaction
13:40and they called it meaningful social
13:42interactions so stuff that got more
13:43interactions was considered better
13:45content even though six months after
13:47they did this change they pulled people
13:49and said hey is your home feed more
13:51meaningful and people said no it's not
13:53it's actually less meaningful
13:55um I there was a single parameter in we
13:59had noticed for at least at that point
14:0118 months when when this report came out
14:03um that a single parameter which was if
14:06you don't just give content credit for
14:08generating interactions like let's say
14:10I'm trying to decide should I show this
14:11to you you know my little virtual model
14:13view says How likely are you to click
14:14How likely are you to reshare How likely
14:16are you to you know put a comment
14:19but you know if you were to re-share
14:21that all your friends might click on it
14:24all your friends might re-share it all
14:25your friends might comment we should
14:26give it credit for for that we should
14:28have a downstream credit the only
14:30problem was they knew that people didn't
14:33like reshares that they've done these
14:35beautiful studies like one of my
14:36favorite graphs from Facebook shows that
14:38you like original content from your 20
14:41closest friends and and Facebook has a
14:43ranking of all your friends they know
14:44which ones are your favorite friends at
14:45least on Facebook
14:46um you like original content from your
14:4820 closest friends but you only like
14:50original reshares from your five closest
14:52friends because people don't agree
14:53shares so content that can actually get
14:56someone to engage after a reshare has to
14:59get over the bias of not liking
15:01researchers and so it turned out to be a
15:03scoring parameter that overwhelmingly
15:05rewarded only really extreme content
15:08and the study came in and said we've
15:09known this for a year and a half
15:11we know that if you take this factor out
15:13you get way less hate speech whales like
15:15literally violence violent content
15:17bloody content violence and citing
15:19content maybe in places that are
15:21considered at-risk countries we should
15:23not include the scoring parameter it
15:26doesn't decrease how long you stay on
15:27the site it doesn't increase decrease in
15:29recessions it doesn't decrease ad
15:31dollars but it will hit our core metric
15:35of meaningful social interactions
15:37which people's bonuses are tied to by
15:39the way and literally in the notes from
15:41Market says with regard to going broad
15:43with taking out the scoring parameter
15:44that we know increases violent content
15:46hate speech all these things if it hits
15:48core MSI we're not going to do it
15:51should they be responsible for that
15:53yeah I think they should yeah
15:56yeah okay then there's the question of
15:57of how much do they contribute to the
15:59genocide but I think you've been
16:00contributing a little bit to a genocide
16:02is something you want to stay away from
16:03or to do it that shamelessly like you
16:06really want a company where you get said
16:07you're not gonna lose any money you're
16:09not losing sessions all the metrics that
16:11are seen externally they're not going to
16:12go down and you know there's going to
16:14say speechless violence less violence
16:16inciting content it's a separate
16:18category you know all those things but
16:20you might have to like figure out a
16:22different way to compensate people and
16:24your CEO doesn't even entertain can we
16:27figure out a way to do it yeah and this
16:30is by the way something I've I've been
16:31reporting on and speaking about for a
16:33while which is the harmfulness of the
16:35share
16:36share it I I think and I'd love to hear
16:39your perspective on this because we're
16:40talking about what I think is the Deep
16:42reshare who your shares have reached oh
16:44no this is just single single single
16:46research okay oh yeah well we don't I
16:48feel strongly about okay yeah we can go
16:50we can go shallow reshares and debris
16:53shares but I always felt that the
16:57ability to share something on impulse
17:01um that's news never leads to good uh
17:05information distribution because it goes
17:08from takes people from being thoughtful
17:10hey I've read this story I think you
17:12should
17:13you know check it out to seeing in your
17:16feed that your political opponent is you
17:18know a space alien Nazi child
17:21killer eater and being like Oh yeah
17:24they're also a blizzard personally
17:25blizzard person yeah and you hit share
17:27yeah and never if you have to put this
17:30under your own identity and copy and
17:32paste that link it can be much more
17:34reticent to do that than to you know
17:37just hit the share button and have
17:39somebody else's you know Avatar show up
17:42in your friend's feed and be like aha I
17:45just did my part for the cause my
17:47parents are both life scientists so my
17:49my mother's a cell biologist my father
17:51is um a pathologist but he studies
17:54viruses and um my mother you know you
17:56can see the the wisdom of my childhood
17:58my mother used to always say be careful
18:00what you select for because you'll get
18:02it or like that's the story of cell
18:04cultures you have to be very careful
18:05about keeping things clean because like
18:07whatever you slept for that's what
18:08you're gonna get
18:09you know when we started talking about
18:11freedom of speech like when the
18:12Constitution was written if you want to
18:14distribute an idea it cost money you
18:17know you had to buy paper you took a lot
18:19of time it took a lot of effort there
18:21was a natural filter
18:22that if you wanted to say something you
18:24had to invest in saying it
18:27um I think asking people to have a
18:29moment of intentionality like a copy and
18:31a paste is not a huge it's not a huge
18:34amount of cost but it adds some costs
18:36you see you don't get
18:38um unintentional sharing yeah everything
18:40yeah and every social network seems to
18:42know this clearly Facebook knows this
18:44the research that you shared showed it
18:46Twitter knew it inherently that's why it
18:49before the 2020 election added the speed
18:52bump that you had to you know do you
18:54want to quote tweet this tweet do you
18:56want to click the link before you
18:57re-share yet they're so addicted to the
19:00virality why I think it's a secondary
19:03thing um so so one of the so I'm finding
19:05a non-profit called be on the screen
19:06because right now we are limited to what
19:08we see on our screens and we need to see
19:10beyond them if we want reform
19:13um you know I think this question of
19:15like why did Twitter do it and Facebook
19:17didn't it's such an interesting question
19:18because it's like a really cheap way to
19:21get like 10 or 15 less misinformation
19:24I think the reason why Twitter did it in
19:25Facebook didn't was Facebook's average
19:27user is substantially less literate than
19:30Facebook than Twitter says I know it
19:31might not seem like that
19:33um but like given the global
19:34distribution
19:36um you know there's
19:37huge numbers of Facebook users that are
19:39literally becoming literate to use
19:41Facebook it is the internet where they
19:43live you know 80 or 90 of all the
19:45content in in the majority of languages
19:48in the world is only on Facebook and
19:50when you look at the outlier countries
19:52there's a really some really good graphs
19:54in there that show you know that there
19:56are countries where 35 percent of
19:58everything that's in the news feed every
19:59impression is a reshare
20:02and so when you have such a large
20:04fraction of your content is just
20:06reshares any friction you add any moment
20:10for pause or intentionality
20:11substantially decreases the content in
20:14the system which decreases consumption
20:16yeah it's a money thing and by the way
20:17if you pull the reshare back or put this
20:22speed bump that causes less resharing
20:24you don't need to engage in these messy
20:27content moderations exactly are you
20:29accused of censors that's my whole point
20:31yeah that's my point since my first
20:33Senate testimony contrary to the
20:35conspiracy theories
20:37I mean it's so obvious yeah
20:40you spend a lot of time thinking about
20:41this so it's obvious it's obvious to you
20:43but it's not necessarily because we
20:45don't we don't study these things in
20:46school it's not really obvious no it's
20:48good to be talking about it in public
20:49inside the companies yeah I mean the
20:52person that that built and I'll get off
20:54my soapbox on this in a moment but I'm
20:56enjoying it so much okay well I'm gonna
20:57keep going then
20:58um the person that that built the
21:00retweet inside Twitter I spoke with him
21:02and he said look it's like I handed a
21:06machine gun to four-year-olds wow
21:08sort of like what it feels like anyway I
21:12I that's my hope is is eventually
21:15because of Revelations like yours and
21:17because of I don't know maybe
21:19conversations like this that have people
21:20inside the company listening we start to
21:22revisit this but you know we may it it
21:25we may end up having product Evolution
21:27before we even need to get there I mean
21:30we were moving to these short form
21:31videos and by the way I'm curious what
21:34you make of this we just had Brandon
21:35Silverman on the show who was the
21:37founder of crowdtangle
21:39um which is this great software tool
21:41that let you see what was moving inside
21:43Facebook public links in particular and
21:45he made a great point that Facebook
21:47actually is de-prioritizing news within
21:50its news feed and you know now we don't
21:54see as many news links
21:56I'm curious what you think isn't is that
21:58something to be spotted uh so part of
22:01part of what frustrates me so much about
22:04um about that is so like the reason why
22:06they're deep deep prioritizing news is
22:09because they're scared rightly they're
22:10getting all this criticism about things
22:11like you know ethnic violence and
22:13they're like they're kind of you know
22:15taking a car from Tick Tock where
22:16they're like you know Tick Tock if
22:17you're not a funny dance you're not like
22:19humorous if you don't like aren't like
22:20like you're not going to get distributed
22:23um and the the thing that's so sad for
22:26me is it really illustrates
22:28you know you can either choose to design
22:31for human scale human modeled
22:34communities
22:35right so I always say like is it a
22:37dinner party is it a cocktail party is
22:39it a church parish is it a conference is
22:41a university you know we have we have
22:43models on how people can exchange ideas
22:45they've lasted for thousands of years
22:48um uh or you end up with AI censorship
22:52and the problem is if you must insist on
22:55content moderation I think you should
22:57have to be transparent about it and that
22:59means things like you should have to put
23:01out samples of the output of your
23:03scoring systems so we can see what's
23:05getting labeled as any of these
23:07categories and what the consequences of
23:09that is because when we take down when
23:13we don't let news get distributed that's
23:16news by the definition of Facebook's AI
23:19we don't know what doesn't get through
23:21or what does get through and when I talk
23:23to activists around the world that work
23:25on issues like
23:27um women like violence against women gay
23:29rights they say across the board our
23:31accounts all get taken down
23:33because these systems that are meant to
23:35keep us safe because they're not done in
23:37a very high quality way counter speech
23:40looks like violating speech and and so I
23:43I 100 am against badly done content
23:46moderation and I think we need to have
23:48some pretty large societal conversations
23:50around like you know if we must insist
23:52on doing content moderation we need to
23:55be doing it in a different way than
23:56we're doing it today
23:57yeah and as many I I feel like as many
24:01system changes we can make before we get
24:04to that decision totally stay up or yeah
24:07that's probably better
24:09so let's talk about another I love how
24:12we're going into product
24:13um I always like I wanted to talk to you
24:15about product for a long time so I'm
24:17glad we're doing it
24:19um another thing that we should talk
24:20about is
24:22um algorithms algorithms on the news
24:25feed and
24:28um just a little interesting history
24:30um I was reading the documents that you
24:33made available to reporters and there is
24:36one internal study that said they
24:38removed the ranking algorithm from the
24:41news feed so and it made it worse made
24:42it worse and he wrote about this he said
24:44I got it wrong yeah I didn't say you got
24:46it wrong yeah yeah did I say you got it
24:47wrong you said that she she didn't see
24:49this documentary she didn't like blah
24:51blah I don't think I said that but I
24:52could be wrong okay anyway we're here to
24:55heal yeah we're here to heal good I'm
24:56all about I'm all about Truth and
24:58Reconciliation so I'm down I appreciate
25:00you sitting down with me and and I I
25:02think that at the time I heard from you
25:04saying that like look um
25:05you're not looking at this explicitly
25:08yeah yeah not yeah you didn't say I was
25:10wrong just you said hey there's some
25:11context you might want to take a look at
25:13and and just so set the set the um the
25:17story straight it it was a study about
25:19what happens when you put the ranking
25:20out where the on the feed which means
25:22that Facebook's algorithm is making a
25:25decision of what you see first as
25:27opposed to the algorithm showing you
25:30um what's happening in a reverse reverse
25:31chronological order and just showing you
25:34the most recently first actually let's
25:36like make that look more concrete for
25:38people who who there's a lot of big
25:39words in there
25:40um so when we open our email
25:42uh the thing that we see first is our
25:45most recent email unless we have some
25:47kind of fancy AI email inbox
25:49um we could have a social network where
25:52we got content from you know things we
25:55chose to get and it showed up you know
25:57in Reverse time ordering just like our
25:58email
26:00um uh and or I say I have a time working
26:02for email but same same kind of thing
26:04like there's different ways we could be
26:05doing this
26:06um and and one of the things that's true
26:08there's a document in the disclosures
26:10that say hey we tried getting rid of the
26:12ranking like the algorithmic ranking you
26:13know where we try to decide what's going
26:15to be better for you and guess what it
26:17has we see more bad things we see more
26:18violating content we see more hate
26:20speech and I think one of the most
26:22important things for people to
26:23understand is that back in 2008 when we
26:27had chronological ranking you know you
26:29saw things in the order they came in it
26:31was kind of like your inbox or reverse
26:33reverse of your inbox so
26:35um when you knew you run out of new
26:37things to see
26:39um we had no big groups
26:42right there were no million person
26:43groups on Facebook Facebook had not been
26:45pushing you for years into groups that
26:48you know you may not have been that
26:49interested in or those groups were
26:51intentionally selected because they were
26:52groups that got a lot of interactions
26:54for contracts like in 2016 they found
26:56out that 65 of people who joined
26:59Neo-Nazi groups in Germany where they
27:01are illegal like you can't have a
27:03Neo-Nazi Facebook group in Germany join
27:05them because Facebook suggested that
27:07group to people
27:08over time though Facebook needs you to
27:11stay on the platform for longer and
27:13longer and your friends and family let
27:15them down they didn't make enough
27:16content for you to get longer and longer
27:19sessions you could click on more and
27:21more ads
27:22we cannot have million person groups
27:25that are basically just like fire hoses
27:27without having algorithms
27:29because that group makes a thousand
27:31posts a day and you flood your feed and
27:34what the article found was if you get
27:36rid of the algorithmic ranking people's
27:38feeds their fees get flooded by these
27:39high frequency groups you no longer see
27:41your friends and family
27:42and so you can't really think about the
27:44idea of taking the current product that
27:46has all these baked in assumptions and
27:48just removing the algorithm you have to
27:49really design design for safety
27:52holistically
27:53and so your point is that if if we were
27:57to go to the most recent first as
28:00opposed to the algorithmically selected
28:02stories that the company would have to
28:05redesign its product it wouldn't be you
28:08wouldn't run that experiment on the
28:09current process exactly you would
28:11rebuild yeah the product which I found
28:13interesting and uh you'd have I think
28:15that's good good context you'd have
28:17groups that were like Discord servers so
28:18you can have a Discord server that has a
28:20hundred thousand people in it and when
28:22the conversation gets too loud you break
28:24off into little subgroups you can
28:25imagine a world where maybe your group's
28:28content never got inserted into your
28:29feed you had to navigate to your group
28:30so you can still have your cancer
28:32support group you just wouldn't have it
28:34insert all the content into your feet
28:36and you're not worried that that the
28:38feed would be overrun by spammers who
28:40just like
28:42post every five minutes and try to get
28:43stuff in front of people so I like to
28:46call the my favorite ranking no no one
28:49else calls it this but I I got to Brand
28:51it because I mean I made it up I call it
28:53algorithm I call it um chronological
28:54plus plus
28:55so in in computer science the way you
28:58add add one if you don't want to write
29:00like n Plus or n equals n plus one is
29:03used to n plus plus
29:05um is uh imagine if you had a little bit
29:08of demotion by frequency so it's largely
29:11chronological
29:12but it if you send 20 posts in a day you
29:16know each Progressive thing demotes you
29:18down a little bit so it's Loosely
29:20chronological and I think the better way
29:22to describe like what I would want is
29:24describable ranking right that right now
29:27we have we have unintelligible ranking
29:29we have a black box neural network and
29:31we don't know what the model really is
29:33optimizing for
29:34you could imagine having a publishable
29:36algorithm like you said it's basically
29:38time based we try to stay as close to
29:40time based as possible but if you post
29:42five times a day if you post three times
29:43a day you know we might move you down a
29:45little bit you'll be less likely to get
29:46seen and that's that's the description
29:48and but there are people who are saying
29:50we should remove protections from these
29:53platforms that use ranking algorithms to
29:56show feed to show you content and even
29:59using cron what did you call it reverse
30:01chronological plus plus it's just a
30:03chronological plus plus yeah you are
30:05ranking in some way but I'm doing an
30:07intelligible way a way that we can have
30:08a conversation on it yeah we can because
30:09right now we never get to have a
30:11conversation yeah on how it's
30:12prioritized yeah I understand yeah and
30:14then but but because that's still going
30:15to be ranking it would still fall under
30:18these Customs reliable so so but my
30:21complaint on on the current way we rank
30:24right is right now we have these ranking
30:26algorithms that are optimized for
30:27business objectives
30:29and you could imagine a world where you
30:31said it's gonna be chronological or it's
30:34me or it's chronological or you have to
30:36give us a defendable reason why improves
30:38safety
30:39right you can't you can't tell us you
30:41made more money on it like that's why
30:42you did it you have to like show us a
30:44thing and show us like what was the
30:45safety objective and you have to do it
30:47publicly and show us consequences so we
30:49can have a conversation on it like
30:51that's a very different world than one
30:53where
30:54um you know they they uh one of the
30:56problems that's outlined the documents
30:57is that safety teams would work for
31:00months and months and months to figure
31:01out a way to make the product safer but
31:03not hurt ads
31:05and because the model was unintelligible
31:07people are just like throwing new
31:10factors at the model and it would often
31:12undo changes that have been made because
31:14you could represent the thing that had
31:15been pulled from the model or the fix
31:16that had been done by lunging together
31:19four or five other features because
31:20people just didn't understand how they
31:22interacted yeah and I like this proposal
31:24yeah so but does that does that then
31:27mean
31:28um companies should should still not be
31:30held liable this what do you think about
31:32this meme because this whole idea of go
31:35completely reverse crime pure reverse
31:37cron without any filtering at all even
31:40explained or not and otherwise you're
31:42liable for what you suggest like another
31:44counter counter offer yeah yeah so let's
31:47imagine instead of like demoting people
31:49who post too frequently imagine if you
31:51posted more than three times a day or
31:53maybe even twice a day your post got
31:55collapsed down to a single line that
31:57said Francis Haugen posted two times
31:59already today do you want to read the
32:01post okay because then it's like a
32:03little shame we're still doing
32:04chronological but you're like hey you
32:06kind of talk a lot I thought about
32:08listening a little you know I was trying
32:10really hard to to have you say that this
32:13um making making platforms liable for
32:16going pure reverse Crown is an
32:18unworkable idea but we got we got pretty
32:20close I think but I like I I think the
32:22way we should talk about liability is
32:24the way we talk about it with cars okay
32:26right where is this question around you
32:28know
32:29um people should have to demonstrate
32:31that they are actively pushing towards
32:32safety and you know if you uh have a
32:36hundred chances at bat
32:38and when we look at you know like one of
32:40the things I've suggested is right now
32:42every time they run an experiment there
32:43are these dashboards and the dashboards
32:46have hundreds of metrics that can be
32:48sliced in a ton of different ways
32:50imagine if they had to publish the
32:52dashboards they don't have to tell us
32:54what the experiment is they don't have
32:55to give up the IP you know they don't
32:57have to tell us about their magic secret
32:58sauce but they have to show us every
33:00single time they run an experiment and
33:01release the data so that we can analyze
33:03whether or not there's a trend where you
33:06know consistently they had chances to
33:08make us safer and they chose not to do
33:09them and they had a whole bunch of
33:11chances they made it more dangerous for
33:12us but it made them more money and they
33:14shipped those like show us what
33:16experiments you run and show us what you
33:17ship or don't ship now if we see a
33:20pattern of behavior
33:22we should hold you responsible for those
33:23consequences I I think that's that's
33:25much better I'm all on team transparency
33:28um and and for context
33:31um I appreciate you unpacking this
33:33um I will link this podcast underneath
33:35that story oh thank you so whenever
33:37someone reads it they'll get you can
33:40read it you can even do it at the top of
33:41the article so they'll actually see it
33:43consider it done
33:45all right uh let's take a break uh we're
33:49here with Francis Haugen she's the
33:50founder of beyond the screen and uh the
33:53person behind all these big Facebook
33:54Revelations that we're speaking about
33:56we'll be back right after this by the
33:58way that was a fake break so um uh we're
34:01going to just insert these dynamically
34:03and take this audio out but if anybody
34:05over here wants to take a 30 seconds or
34:0845 without breaking the leg and get a
34:10seat why don't we just do that and then
34:13we'll pick back up in the second half
34:15and if you have to leave I won't hold it
34:17against you it's late on a Friday yeah
34:19questions
34:22my pleasure I I you've never asked me
34:25too hard a question and I will never
34:26feel attacked okay awesome I I really
34:28believe
34:30um the way the world changes is one
34:32conversation at a time yeah and I
34:34guarantee there are Skeptics out there
34:36and you know the way we change Hearts is
34:38the Skeptics get to ask questions and so
34:40I appreciate hard questions awesome well
34:42thanks for rolling with it
34:47all right everyone it's okay if there's
34:49a little background noise I'll just keep
34:50going but don't feel bad about uh making
34:52you don't you don't change the world by
34:54converting your your fans you change the
34:57world by converting your Skeptics yeah
34:59I hear you yeah okay
35:02and we're back here for the second half
35:04of big technology podcast we're here
35:06with Francis Haugen at unfinished live
35:08by some miracle or actually probably the
35:11draw of what Francis has to say our
35:13second half audience is actually larger
35:15than our first half audience
35:17substantially larger much bigger this
35:18has never happened uh for a podcast that
35:20I've done before live so
35:23um anyway let's hear it from you guys
35:26thanks for for showing up because it's
35:27got to be loud
35:34thanks everyone for being here
35:36um all right let's let's get into the
35:37second half
35:38uh one interesting thing you said in the
35:40first half
35:41um that that caught me by surprise was
35:43that Tick Tock acknowledges its power
35:45and Facebook Facebook doesn't
35:47really yeah I mean like Tick Tock is
35:49designed to be censored right like we've
35:51had lots of scandals around this right
35:54so
35:54um how many people remember uh I don't
35:57know it was a couple years ago there's
35:58like a makeup tutorial where she's like
35:59this is how you do a smokey eye and like
36:01blah blah oh and by the way the uyghurs
36:03are being killed does anyone remember
36:05that I do yeah we're good
36:07um so so the reason why she did the
36:10smokey eye for like 15 seconds and then
36:12started you know spilling the tea on you
36:14know they're killing a million people
36:16you know or they have a million people
36:17in concentration camps is um they were
36:20tick tocks manually sensors all their
36:22highly popular content and
36:24um you know I don't agree with any
36:27system where they take down lots of
36:29content and they don't tell us why or
36:30how or they don't show us the biases and
36:32what they take down or don't take down I
36:34highly don't support that but they
36:36acknowledge the idea that social media
36:38has power and the social media has
36:40social implications
36:41and
36:43um I
36:45and and I think the fact like they've
36:46publicly talked about the idea that they
36:48watch for people who fall into rabbit
36:49holes and they try to pull them away
36:50from those do they actually or not I
36:52don't know they're not transparent
36:54either
36:55um but they understand the idea that
36:57they have an impact in society and so
37:00um are they rooting for our team I don't
37:02know
37:03but acknowledging power I think is
37:04better than than pretending you know the
37:06World is Flat and you know everyone's
37:08the same and you're just a mirror and
37:10everything bad has always been bad and
37:12like now you can see it and that's why
37:13you're mad if you rather have a platform
37:15acknowledge power but
37:18um actually like work to to do things
37:21that undermine our society or one that
37:23doesn't acknowledge its power but
37:25undermines our society he's trying I
37:28don't know I I I would I would be um I
37:31would be highly hesitant to say that
37:33Facebook is trying right like like
37:36Facebook
37:37um if Facebook earnestly was trying
37:41you know when when they got asked please
37:44share the research with us that you have
37:45internally on whether or not kids are
37:47killing themselves because of your
37:48product
37:49they would not have given back a
37:51three-page letter saying what research
37:54yeah they would turn over their research
37:56they lied to the oversight board also
37:57yeah like blatantly blatantly so for
38:00context for people
38:01um uh when they they asked the oversight
38:04board hey like you know you know was
38:06taking Trump down the right thing can we
38:07have a conversation about it
38:09um the oversight board came back with
38:11um like 15 or 16 questions and and they
38:14were on a variety of things and Facebook
38:16wrote back with answers and um one of
38:18them was uh they asked like hey this
38:21this uh crash check program this idea
38:23that VIPs get a special pass like you
38:25say you treat everyone the same but then
38:27you also say like the reason why Trump
38:29hasn't gone touch is because like he's
38:31in crosscheck he has more context on
38:33that and they said oh we've already
38:35explained what Crosstrek is and they
38:36linked to a blog post from 2018.
38:39and that blog post said things like this
38:42is not a two-tiered system this is just
38:44like getting a little extra double check
38:46to make sure our policies are accurately
38:48being reinforced
38:50and in reality
38:52because they weren't willing to spend a
38:54very small amount of budget
38:57they didn't actually put that second
38:59check in for for most Safety Systems so
39:02for most Safety Systems people got white
39:04listed so if you were in Cross check
39:05those safety systems no longer applied
39:07to you
39:08and even for the ones where I did apply
39:10because they were they were willing to
39:11allocate so little budget to this this
39:14VIP program uh it would take like three
39:16days before a human would look at your
39:18content and so people would get about 75
39:20of all the distribution they would have
39:22gone anyhow because of that on your
39:24funding
39:25okay what do you think of the oversight
39:27board I mean the the meme about the
39:30oversight board among Facebook critics
39:31is that it's a
39:33public relations stunt that's a shield
39:35that you know will make it look like
39:38Facebook actually cares what the public
39:39has to say but it doesn't do you
39:41subscribe to that I
39:43deeply deeply deeply manipulative so so
39:46if we were to ask like what does the
39:49Facebook oversight board do you know if
39:51you believed in the theory of what's in
39:53what what's on the labels in the can
39:55you'd be like oh I mean they they do
39:57oversight you know they they look after
39:59Facebook as a whole it really should be
40:02called the content moderation appeals
40:04board
40:05because that is the entire scope of what
40:08they can do they can't ask about the
40:10performance of the content moderation
40:11systems they can't even ask like what
40:13language are the content moderation
40:14systems in they can't ask about how
40:17systemic or like how some are there
40:18other examples of the ones that get
40:20appealed to them they can't do anything
40:21they can't ask about the algorithms in
40:23general
40:25um and so I my heart really goes out to
40:26the oversight board because it is full
40:28of a number of very conscientious people
40:31who I think genuinely try very hard but
40:34because they are
40:36um so limited by what Facebook allows
40:38them to do
40:39um uh they they get put in a very hard
40:41position what do you think about the
40:43metaverse I
40:46all the murderers
40:48um so the metaverse will will make
40:50certain problems easier and some
40:52problems harder right so on the side of
40:55um reducing risk
40:57um you know we talked about before your
40:59friends and your family are not the
41:00problem on Facebook
41:02hyper virality million person groups
41:05five million person groups in a lot of
41:07African countries the single largest
41:09news source in the country or maybe the
41:11the only really Mass news source will be
41:13like a five million person Facebook
41:14group
41:15um
41:16that's the problem with misinformation
41:18that we have a a selective fire
41:21megaphone where a thousand posts come in
41:24but only two get delivered to you and
41:26they're the most extreme too right
41:28that's the problem in the metaverse it's
41:30much more about one-on-one interactions
41:32or small group interactions and so
41:34you're going to see probably less things
41:36less things like misinformation problems
41:39the big problems with the metaverse are
41:41things like um you know studies out of
41:43China have shown it's more addictive or
41:45more habit for me
41:46um like substantially more so than
41:48phones because it's immersive I think
41:50there's gonna be a bunch of
41:51psychological things that we're going to
41:53be unpacking over the next few years
41:54things like you have a socially isolated
41:57child who
42:00um you know maybe they are of a lower
42:01socioeconomic tier they can't afford
42:03after school activities maybe they have
42:05a single parent
42:07um they they don't really like their
42:08life and they come home and they put on
42:10their headset and they have a way cooler
42:12apartment their friends are more
42:14beautiful have cooler clothes they do
42:16these fun things but at the end of the
42:18night they take their headset off and
42:19they look in the mirror and they no
42:20longer let they don't like the person
42:22they see
42:23right the idea that you would the only
42:25person you liked being was a virtual
42:27Avatar I think is just it's it's
42:29heartbreaking to me
42:32um and I worry that the metaverse is
42:33going to become the stop gap for us
42:35where we say instead of like making sure
42:37our old people are our our elders are
42:40like integrated into our communities and
42:41have adequate resources we can like plug
42:43them into the metaverse and leave them
42:45alone or instead of building community
42:47centers and after school activities we
42:49can you know leave our kids to to go
42:51into their and so I worry a lot about
42:53these kind of systemic problems and we
42:55we know we know that they haven't
42:57invested in safety again so when it
42:59comes to things like hate speech
43:01we're creating these spaces where you
43:03know if you can hear 40 people speak but
43:05you can't you know pause the tape and
43:08roll back two minutes and realize the
43:10person who's been shouting racial
43:12epithets at you is that one you
43:14basically can silence the ability of
43:16minorities or women to be in these
43:18spaces because they just get yelled at
43:20and they get pushed out and so I think
43:22there there should be upfront investment
43:24in safety and shared systems across the
43:26games because it is very clear that the
43:28reason why Facebook wants to go to the
43:30metaverse is they want when someone
43:33comes to them and complains about
43:34getting groped or complains about
43:36getting you know harassed that that they
43:40can say I'm so sorry that's that's
43:41really tragic but but we didn't write
43:43that out we just made the hardware
43:45so you think it sounds like you think
43:46the metaverse is going to happen because
43:48that's still an open question
43:50I am a a follower of Moore's Law so I
43:53think the headsets are going to get
43:54cheaper I think they're gonna get faster
43:56lighter they're a higher resolution
43:59um I I I I have trouble imagining that
44:0310 years from now we won't spend more
44:04time than we currently do in the Bible
44:05verse do I think Mark's vision of
44:08spending 12 hours a day in the metaverse
44:10will come true no I think Mark does that
44:13Mark thinks that because Mark lives that
44:16but not all of us have to avoid cafes
44:19because people glare at us when we walk
44:20in
44:21so so you know we have different
44:23incentives
44:24another reason why Facebook wants to go
44:26to the metaverse okay maybe there's the
44:28you know we don't need to be responsible
44:30but the real reason is that they are an
44:33app on Apple's phone app on Google's
44:35phone they want to own the phone they
44:37want the operating system no doubt about
44:39that
44:40um
44:41an interesting thing has happened where
44:43Apple has asked its users if they want
44:48to cut off Facebook's ability to track
44:49them off the app and people have
44:51overwhelmingly said yes yep now an
44:55interesting thing has happened after
44:56that where apple is in the middle of
44:58doubling its ad sales team or its
45:00advertising team not sales the team that
45:02builds ad products there
45:04I gotta ask you about this if the
45:06business model drives
45:08the problems that we're seeing and apple
45:11is kneecapping Facebook's business and
45:14in some ways emulating it
45:16don't they put themselves at risk of
45:19falling into some of the same problems
45:21that we see Facebook facing today
45:24such a great question
45:26um I do think that they are playing with
45:28fire like I think it's always hard when
45:31you when you have an economic incentive
45:33that is not so right now Facebook's or
45:36Apple's economic incentive is safety
45:38right they make their money selling new
45:39devices so if you get more privacy and
45:42having more privacy it makes you want to
45:43buy their phone more they can make more
45:45from your phone
45:47um I think they are uh they're they're
45:50mixing their incentives and and I have
45:52not sat down and like fully wargamed out
45:55that but I I I
45:56so I would want to be cautious to speak
45:59with too much confidence but I can
46:02totally imagine that might become a
46:03problem in the future I think it's a
46:05major problem for them and we're just
46:07scratching the surface there yeah it's
46:09definitely worth talking about yeah
46:11um let's talk a little bit about your
46:13your non-profit
46:14um I want to ask this question
46:16beforehand I read this interesting quote
46:18from you that said like
46:20um your greatest hope is that you're
46:22you're not relevant I won't be needed
46:23anymore not needed anymore but at the
46:26same time you're working on a book you
46:28sign with CAA which is a big
46:30Hollywood agency so
46:32how do you square those that your hope
46:34and the action well I I right now I
46:37don't see very many people who are on
46:39the stage who are proposing and pushing
46:41for things like you know let's solve the
46:43product not the content
46:46um uh at the same time I'm someone who's
46:47almost died right like I spent two
46:49months in the hospital when I was 29.
46:52um I am fully sensitive to the idea that
46:54like life is fleeting and precious and
46:56the way we solve real problems in ways
46:58that have the highest chances of of um
47:02of working is you cannot be reliant on
47:04individuals right we have a hit by the
47:07bus problem right now like we should be
47:09wary of like Facebook employees driving
47:11buses around New York right like you can
47:13imagine nothing that would be more
47:15convenient for them
47:16we need a hundred thousand people who
47:19can speak as detailedly about the
47:22choices we have with regard to social
47:24media as I can
47:25right a world where there's only one
47:27person talking this way is not a safe
47:30world
47:31and I I think anything I I had a lot of
47:34people pushing me really hard when I
47:36first came out to be like give us the
47:37five-point plan how do we fix Facebook
47:40um and I I have said repeatedly any
47:43world where you know I am the solution
47:46is just as autocratic as the world we
47:48live in right now with Facebook and and
47:50I really believe that the process
47:52process of democracy is about having a
47:54lot of different people with different
47:56interests and different stakeholder
47:57groups
47:58having to hash out something that
48:00they're they're willing to feel
48:01comfortable with together and right now
48:03the people who get to sit at that table
48:05are very very small number of people
48:07like there are hundreds of people in the
48:09world right now who understand these
48:11systems at the level of detail that I do
48:13right the interaction between the
48:15algorithms and the product features and
48:16how all these things feed back on each
48:18other how they differentially impact
48:19different kinds of people and when
48:22um and that is terrifying to me
48:24and so we need a world where there is
48:27not and when you think about a Continuum
48:28of expertise like we need peop we need
48:32um right now it's not just that the only
48:34security people get trained inside the
48:37companies right it's as if SpaceX was
48:39dependent on training as aeronautical
48:41Engineers in-house
48:42if SpaceX was training its aeronautical
48:44Engineers in-house we would not be going
48:46to Mars
48:47period
48:49the reason why Twitter can't afford
48:50safety people
48:52is because there's so few of them and
48:53Twitter can't pay high enough
48:55all of us are endangered by the fact
48:57that we are reliant on these companies
48:59to train the experts but if we look on
49:01the policy side it gets even more scary
49:03because just like those security people
49:05are trained in-house
49:07the only place where you begin to become
49:09fluent in these policy issues around
49:11social media is like working on these
49:12issues on the hill you're a legislative
49:14Aid your Congress person really cares
49:16about it you start learning more about
49:17it you get more fluent and as soon as
49:19you start to be a little effective you
49:21get poached off by Facebook you get
49:23poached off by Twitter you poached off
49:24by another tech company
49:26and in the the couple weeks before my
49:29Senate testimony we met with an aide
49:32from blumenthal's office and one from
49:33Blackburn's office and shared with them
49:36we'd given them the SEC filings and all
49:38the files associated with it and we
49:40asked them like hey like you know um
49:42this is a really good conversation is
49:43there anyone else on the committee that
49:45would benefit from having these
49:47documents
49:48and I don't remember which eight it was
49:50but one of them was like yeah you know I
49:52that I I get why that would make sense
49:54and I want to say yes
49:56and I would say yes if we could go more
49:59than two weeks without having a staffer
50:01from the committee leave for Facebook
50:04two weeks
50:06right and that's a huge problem did you
50:08see it this just happened but there it's
50:10a revolving door problem across
50:12government they've hired people from the
50:13FTC there was a banker that was uh no a
50:17banker Bankers were testifying before
50:18Congress and um one of the members
50:22thanked the bank for hiring one of their
50:25staffers oh the bank was like her father
50:28already works here it's amazing
50:31it's a it's a problem the revolving
50:33doors is a problem no problem
50:35um when it comes to Banks like you know
50:37you can you can get a finance policy
50:39degree like if you want to go get your
50:41PhD and specialize in like how should we
50:43regulate Banks you can go and and get
50:45probably a free degree in public policy
50:48and that's in the subject you can
50:50probably get grants for it
50:51um you cannot take a single class
50:55single class where you really understand
50:58here's the spectrum of choices we have
51:00you know here's the very wide eclectic
51:03palette of colors we could paint with to
51:05build social networks and here's the
51:06consequences of these choices you cannot
51:08take a class if you're a policy person
51:10or even a CS person anywhere in the
51:12United States right now anywhere in the
51:13world
51:14and that's unacceptable
51:16and I've heard you talk about the fact
51:18that you want to build a simulated
51:20social network so people can train on it
51:22you also want to train lawyers to who
51:25are interested in bringing action on the
51:27ins and outs of these networks so they
51:28sound smart about it is this what beyond
51:30the screen is going to be all about so
51:31beyond screen is two core products or
51:33two core things we're working on one is
51:35a project that we got seed funding for
51:37and we announced yesterday so the
51:39McCourt Institute is graciously giving
51:40us our first seed funding for something
51:42around duty of care so when we come when
51:45it comes to any other physical product
51:46any food you eat any drug you take we
51:49have certain expectations about what the
51:51floor is for safety like what are the
51:53what are the questions you should ask
51:54like you should have done a certain
51:56amount to like make sure this was a safe
51:58thing we don't have an expectation
52:00around that we don't have a consensus
52:02around what that would be yet when it
52:04comes to social platforms and I say
52:06social platforms because I would also
52:07include things like games or Discord or
52:09things like that in there and um so
52:11we're doing a project that's around uh I
52:14think Jonathan height has done a really
52:15good work around helping us begin to
52:18have a public conversation about the
52:20harms around kids Mental Health
52:22we want to mirror that process across a
52:24wider set of harms let's talk about
52:27human trafficking let's talk about
52:28cartels in addition to kids let's talk
52:30about Elder issues you know let's do a
52:32much broader set and reproduce that
52:34process let's kind of do the Truth and
52:36Reconciliation phase
52:37and and because I'm a product manager
52:40like the way we make solve problems is
52:42we get really clear about articulating
52:44what the problem is and then let's talk
52:46about levers that would prevent or
52:49mitigate that problem
52:50because often right now we confuse a
52:52lever for strategy for pulling that
52:55lever I'll give you a concrete example a
52:58lot of the harms for kids a common lever
53:00is let's keep under 13 year olds or
53:02let's keep under 16 year olds off those
53:04off the platforms
53:06um right now because people who
53:08understand social problems often don't
53:09understand Tech
53:11they reconverge on list check IDs
53:15and I don't think checking IDs works
53:17like you can go like we can have a
53:19conversation on it I also think there's
53:20huge Collateral Damage I don't think any
53:22of you really want to have to have your
53:24ID checked there's like a variety of
53:26issues with it but if you had asked a
53:28technologist hey I have a problem I need
53:30to find under 13 year olds how could I
53:32do that they would come back and say
53:33Here's 10 or 15 ways to find it under 13
53:35year old we don't have to check IDs if
53:39we could all look at the same list and
53:41say hey here's the menu of what's
53:42possible Right now the platforms are
53:44only doing one or two things they're
53:46like asking you how old you are and if
53:48you say you're 13 you're like cool
53:50you know I'm not saying if you all 15
53:52things like I don't want to be
53:54prescriptive but I do want us to sit
53:56down and say hey maybe there's six or
53:57seven things that we all feel
53:58comfortable about that we think are the
54:00expectation and if you're not willing to
54:02do that amount of work then you're
54:03negligent so that's the goal of Duty of
54:05care and then because a lot of the logic
54:08and duty of care is actually dependent
54:11on having an intuition on the physics of
54:14social networks instead of saying just
54:16trust me you know like we talked about
54:18deep researchers versus shallow reshares
54:20or like how long should you have to wait
54:22before you re-share all these things
54:23imagine if we had you know we spent
54:25hundreds of millions of dollars every
54:27year on flight simulators to train jet
54:30Pilots to keep us safe for the Air Force
54:32or keep our Jets safe at you know for
54:35Delta Airlines
54:37um we spend zero dollars on flight
54:40simulators for social networks we have
54:42to train all of our safety people
54:44in-house because we don't have a lab
54:47bench for training them in Academia
54:50and so those are the two things that we
54:52want to work on over the next few years
54:54sounds great I read that you're trying
54:57to raise or you were trying to raise
54:59five million dollars for this
55:01um I have a couple questions about that
55:02a I'm curious if you've raised it b
55:06um if you have where's it come from and
55:08I would say most importantly there are
55:11there's a I think a plague in this world
55:14a big Tech advocacy where there's so
55:17much dark money that gets spent all over
55:20the place and no one ever discloses
55:23where that money is coming from and so
55:25we don't know where the agendas lie so
55:27can you I don't know if you have or not
55:29but are you going to commit to share the
55:32list of your funders moving forward when
55:34it comes to running this non-profit you
55:36know I I I am totally willing to
55:39entertain that idea I want to put a
55:40quick caveat which is I would like to
55:41check in so we've only taken we've only
55:43taken two checks ever likely McCord is
55:46we're in the process of getting the
55:47check from them but they made a public
55:49announcement of it yesterday I would
55:51want to check with those two other
55:52funders on whether or not they would
55:53want their names publicly but for
55:55context prior to my court we'd only
55:57raised
55:58um on a very low
56:03um six figures
56:05um like I've I've gone by on a
56:07shockingly small amount of money and
56:09part of my book Advance is all that has
56:11funded me since I came out
56:13um the people who tell me oh it's easy
56:15they'll just give you a check for 300
56:16000 I I don't know how that actually
56:18works because it has never happened for
56:20me
56:21um not gonna complain though
56:23um but I I like that idea and I can I
56:26want to double check first but we've
56:27been public about our funding at least
56:29in the last
56:31uh are after those first two checks I
56:34feel like you should make it a
56:35prerequisite and maybe that's the
56:36question that I kept going forward I can
56:38talk to them and be like hey like we
56:39need to have your name out of the open
56:40So eventually fine with that um oh we're
56:42running out of time
56:43do you want what 10 more minutes or are
56:45they going to get angry at you we're the
56:47last ones so are you guys mad
56:49they're mad
56:50um let's do five more and then and then
56:52we can be Shameless we'll make them get
56:54pull us off the stage yeah sounds good
56:57um but this is just a curiosity of mine
57:00um in the 60 Minutes interview you did
57:02after you revealed uh who you were
57:05um maybe it was there maybe it was the
57:07journal you you mentioned that
57:09um you you had basically been living off
57:13of Bitcoin winnings or something like
57:15that I don't know correct the record if
57:17I'm wrong I don't think there's anything
57:19in 60 minutes oh okay but was is that
57:22true because you were I heard you were
57:23living in Puerto Rico no I think people
57:25could sleep yeah yes yeah they complete
57:27the two so I I do hold some crypto it is
57:30not a life-changing amounts of crypto
57:32okay um it is part of a diversified
57:33portfolio all right
57:35um uh but I live in San Francisco so
57:37like I have lots of people who talk
57:39about crypto all the time
57:41um uh I moved to Puerto Rico um because
57:43I I uh when I I was so ill that I was
57:47paralyzed beneath my knees so I I went
57:49into hospital because my leg turned
57:51purple because I had a foot long clot
57:52that had been in my leg for chain 18
57:54months in two years they don't know how
57:56long and when I went to the doctor and
57:58complained about it
58:00um I'd be like I'm in a lot of pain I'm
58:01really exhausted they'd be like oh it
58:03sounds like you have depression you know
58:05there's nothing gendered about women's
58:07pain
58:08clearly but it turned out that I was
58:11also starving to death so I had celiacs
58:13and I was
58:14um even though I gained weight
58:16I was so malnourished my protein
58:18synthesis had stopped so I showed up in
58:21the hospital and and for context my
58:23father-in-law died last week and he had
58:26been suffering from cancer for a long
58:28time maybe seven months and he went
58:30through a period of time where he didn't
58:31really eat anything for four months he
58:33had very very little and he lost 75
58:36pounds or something insane and and
58:40my father was reassuring me as he was
58:43looking through his lab results he's
58:44like these look real bad but to be fair
58:46your numbers looked worse
58:49um and uh so I was really malnourished
58:51and I was I got paralyzed beneath my
58:54knees
58:55um Alex my nerves my nerves died and I'm
58:58much better now I can walk on my toes I
59:00hide for hours but I have constant pain
59:03I have constant very severe pain
59:05um to the point where in addition to my
59:07neuropathy drugs I have to numb my feet
59:08like that level of pain because I can't
59:11take more because my heart beats too
59:13slow for context and um uh I was really
59:17unhappy in San Francisco during covet
59:19because if you've ever been you know how
59:21balmy and warm San Francisco is
59:24um I used to live there yeah yeah and
59:26and the buildings aren't insulated so
59:27like I couldn't even keep my inside warm
59:31um and I I was really unhappy and a
59:33bunch of people were like we're all
59:33gonna move to Puerto Rico and I was like
59:35well we have to only socialize Outdoors
59:38because it's covered
59:40you know I I have nothing to lose and I
59:42went down there for like a three week
59:43vacation and I was like wow
59:46I've been complaining about being cold
59:49for years why did I live in San
59:51Francisco
59:53um and I it's just been a game changer
59:55it's I I actually credit Puerto Rico
59:57with my ability to have been as
1:00:00successful as I was with the whistleblow
1:00:02because I had so much more energy after
1:00:05I moved there right I was able to lower
1:00:07my pain meds a bunch
1:00:09um and it it made it um
1:00:12it just it gave me many more hours a day
1:00:15where I was effective okay so we get to
1:00:17the bottom of the crypto narrative we're
1:00:19out of time do you have like 60 seconds
1:00:22is there anything you're optimistic oh
1:00:24sure can I can I so I can can I do my
1:00:26soapbox now yeah but they're gonna make
1:00:28us rap but yeah definitely so bucks away
1:00:30I'll hold the floor okay um Okay so
1:00:34we talked about some heavy stuff today
1:00:35right so like genocide like you know
1:00:38business incentives that are dystopian
1:00:40kill democracy whatever
1:00:43um the reality is every single time we
1:00:46have invented a new communication
1:00:47technology it has been really really
1:00:51disruptive like people like to say the
1:00:53printing press was disruptive the
1:00:55printing press was not disruptive
1:00:56because no one could read right it's
1:00:58like 60 years between the invention the
1:01:00printing press and the printing press
1:01:02causing harm the real thing that caused
1:01:04harm was Martin Luther was like the
1:01:06Catholic church has been lying to you
1:01:08you're not going to hell just go be nice
1:01:09to people love each other let's teach
1:01:11everyone to read so they can read the
1:01:13New Testament too and it turned out when
1:01:15you taught people to read it went from
1:01:17being three percent litter in Germany
1:01:19being 30 litter in 20 years and people
1:01:21started publishing pamphlets on things
1:01:23like you know how do you know if your
1:01:24neighbor's a witch
1:01:26you know I hate to tell you you might
1:01:28have to burn them it's gonna be hard but
1:01:29you can do it
1:01:31um we invented cheap printing presses
1:01:32those are newspapers and we literally
1:01:34had Wars over misinformation
1:01:37um uh historians have looked at the
1:01:39telegram Telegraph Telegraph the
1:01:42telegraph did not cause the Civil War
1:01:44but they believed that the telegraph
1:01:46influenced when the Civil War was
1:01:48because For the First Time people could
1:01:50could actually see that the opinions in
1:01:53DC the opinions in other parts of the
1:01:55country were very different than the
1:01:57ones at home because the news all
1:01:59converged into real time
1:02:00uh radio we've had whole genocides over
1:02:04radio things like Rwanda or like World
1:02:06War II for the first time people had
1:02:08personal relationships with their
1:02:09leaders and and and literally Hitler was
1:02:12the best person in the world at Radio
1:02:13what we are experiencing right now is
1:02:16not novel
1:02:18but the reason we feel overwhelmed why
1:02:20this feels different why is this so
1:02:23scary
1:02:24is because it's actually our problem to
1:02:26solve
1:02:27right like we feel overwhelmed because
1:02:29we can see that we're in the driver's
1:02:30seat that we have to figure out the way
1:02:32forward
1:02:33and I want to be honest it might get a
1:02:35lot worse before it gets better
1:02:37but the reality is is we have figured
1:02:39out how to do this every single time
1:02:40before and I have a hundred percent
1:02:42confidence
1:02:44that if we do bring a hundred thousand
1:02:46if we bring a million more people to the
1:02:47table we are going to figure out a path
1:02:50forward where we get to live with social
1:02:52media where we get to feel good about it
1:02:54and it's good for us so I'm excited
1:02:56about the future Francis Haugen everyone
1:03:01thank you
1:03:08thank you Francis for the chat thank you
1:03:10what an amazing audience you guys are
1:03:12amazing
1:03:13um this is gonna be on big technology
1:03:15podcast feed so if you go on your phone
1:03:17subscribe to Big technology podcast it's
1:03:18available on all podcasts app would love
1:03:21to have you we do these every single
1:03:22Wednesday uh We've interviewed francis's
1:03:25lawyer
1:03:27um or lessig oh Lawrence yeah yeah he's
1:03:29wonderful and uh and a bunch of other
1:03:31people that agree and disagree with her
1:03:33so thanks again thank you to unfinished
1:03:35live uh for having uh me here having
1:03:37this conversation with Francis
1:03:40um it's been awesome and uh well we hope
1:03:43to see you again on the internet
1:03:47thank you

About this transcript

This page contains the full transcript of Facebook Whistleblower Frances Haugen On Product, Competition, And Transparency by Alex Kantrowitz, generated from the public captions YouTube serves with the video. The transcript has 12,505 words across 1,852 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.