[00:27.450 --> 00:28.400]  Welcome back.
[00:28.760 --> 00:30.640]  Is there hope or is there hope?
[00:34.680 --> 00:35.840]  We're doing this.
[00:36.520 --> 00:39.860]  The impossible has happened yet again here in New York City.
[00:40.180 --> 00:41.740]  Thanks so much for being part of it.
[00:42.040 --> 00:48.000]  Many of you were here during the opening ceremonies yesterday when we talked about being excellent to one another.
[00:48.140 --> 00:50.080]  We talked about having a great time.
[00:50.440 --> 00:54.420]  We talked about how wonderful the program is and has all that come true or what?
[00:55.100 --> 00:55.540]  Right?
[00:56.720 --> 00:57.500]  It has.
[00:59.880 --> 01:13.440]  On behalf of some of the organizers, speaker committee, some of the other groups that I work with, it's been such a thrill to see it come together and such a thrill to see how, as always, it's made up by you, the audience, the people that participate,
[01:13.640 --> 01:22.700]  the people that engage, the people that volunteer and help out, all the speakers, the people that are doing workshops, different types of presentations, art, vendors, contests.
[01:22.800 --> 01:23.380]  It's just amazing.
[01:23.820 --> 01:25.540]  It's so much of a bottom-up effort.
[01:25.660 --> 01:26.380]  And this is why we're here.
[01:26.480 --> 01:27.240]  This is what we're about.
[01:27.480 --> 01:29.560]  So thanks to everyone for being part of it.
[01:29.740 --> 01:35.480]  I hope you're not too tired yet because we're only really halfway through the program at this point.
[01:35.920 --> 01:41.760]  Just to give you a little bit of a highlight, I'm giving a slightly longer intro because we have an hour and 50 minutes maximum today.
[01:43.880 --> 01:45.280]  Just a little bit of what's coming up.
[01:45.420 --> 01:48.520]  First of all, tonight, hackers got talent in this room.
[01:49.020 --> 01:50.860]  No signups necessary.
[01:51.060 --> 01:54.160]  Just show up with your talent and show us what you got.
[01:54.300 --> 01:56.320]  Jason Scott will be in the house.
[01:56.660 --> 02:03.080]  I'd also like to mention, you might have heard about this already, we did have a schedule change 10 o'clock tonight.
[02:03.600 --> 02:06.980]  The steganography talk will not be occurring at 10 o'clock tonight.
[02:07.180 --> 02:09.100]  That's, we hope, going to occur tomorrow.
[02:09.320 --> 02:10.420]  There's some travel issues.
[02:10.900 --> 02:14.480]  10 o'clock tonight will be a medical device hacking.
[02:14.820 --> 02:19.460]  And that is focused on someone that a lot of us have seen over the years.
[02:20.240 --> 02:21.380]  Someone named Jim.
[02:21.880 --> 02:28.540]  Jim is someone that has a whole story of medical history, is in dire health at this point.
[02:28.660 --> 02:32.460]  And this is more or less a story of how they were hacking medical devices.
[02:32.620 --> 02:33.360]  So that sounds pretty cool.
[02:33.480 --> 02:35.140]  So I just want to mention schedule change.
[02:35.340 --> 02:35.980]  10 p.m.
[02:36.140 --> 02:37.260]  tonight, room 206.
[02:37.860 --> 02:39.140]  Hacking medical devices.
[02:39.380 --> 02:41.140]  He's not dead yet, Jim.
[02:43.420 --> 02:47.060]  Slightly bad taste, but that's the title we got.
[02:48.100 --> 02:53.120]  And steganography, hopefully tomorrow at 10 a.m., but to be determined.
[02:54.700 --> 02:58.600]  So we do still need also a few volunteers.
[02:59.120 --> 03:01.260]  There's still some open security shifts.
[03:01.580 --> 03:11.060]  We are looking for a workshop helper, someone that would assist Mitch and the various workshop participants with setting up workshops, transitioning workshops, supporting workshops.
[03:11.220 --> 03:13.540]  That would be, that would also be very welcome.
[03:13.700 --> 03:19.880]  So a couple other announcements that, that you might have seen on the wiki as well concerning fourth track.
[03:20.020 --> 03:21.640]  So we have some fourth track talks.
[03:21.800 --> 03:22.700]  Those are on the wiki.
[03:22.840 --> 03:27.520]  They're going to be in the InfoBeamer rotation starting at the next hour as well.
[03:27.740 --> 03:32.120]  And that's going to be, that's going to be a couple of pretty interesting talks.
[03:32.360 --> 03:44.000]  If you want to give a fourth track talk, which means that you're not on the program, you didn't, you know, uh, uh, get into the formal program, but you have something to talk about, sign up at the InfoDesk.
[03:44.200 --> 03:46.980]  And, um, and then the InfoDesk will put it in the wiki.
[03:47.480 --> 03:51.500]  The MCs will try to read it in between talks, you know, what's, uh, what's coming up.
[03:51.620 --> 03:57.360]  And then you can give your, uh, your talk down, and this is in the coffeehouse on the, uh, on the third level.
[03:57.360 --> 04:02.000]  So that's a little bit of a summary for me, a little bit of what's going on, but mostly it's to say thank you.
[04:02.220 --> 04:02.820]  It's working.
[04:02.820 --> 04:03.640]  We're here.
[04:03.840 --> 04:04.460]  It's happening.
[04:04.660 --> 04:16.920]  And it all is credit to all of us for coming together and being patient and supportive and engaged and interested and having some incredible, incredible content.
[04:17.080 --> 04:19.640]  So let's give us all a hand.
[04:26.020 --> 04:26.560]  All right.
[04:26.820 --> 04:32.360]  So, um, as I said, I wanted to give just a few extra remarks because we had a little bit of extra time in this session.
[04:32.500 --> 04:35.040]  We're really thrilled about this, uh, upcoming talk.
[04:35.240 --> 04:39.540]  This is a keynote discussion and Q and A with, uh, with, uh, Sophie Zhang.
[04:39.880 --> 04:45.780]  We're going to have, um, Yan Zhu as well, who is, um, the person going to do sort of the interview.
[04:46.020 --> 04:49.800]  And they were, uh, on the stage with Chelsea Manning a few years back.
[04:49.980 --> 04:51.580]  Some of you might, uh, some of you might recall.
[04:51.980 --> 04:53.900]  The, uh, the story is what we're going to hear.
[04:54.400 --> 04:57.280]  Uh, Sophie was a Facebook whistleblower.
[04:57.440 --> 04:58.720]  We're going to hear some insights.
[04:58.720 --> 05:06.520]  We're going to have some discussion and, uh, this will, this will, um, take place for, uh, a little while, maybe an hour or so.
[05:06.640 --> 05:09.480]  And then we'll also have opportunity for questions from the audience.
[05:09.660 --> 05:14.720]  So if you do have questions, uh, try to hold them until, until we open up for questions.
[05:15.260 --> 05:23.760]  And questions as usual will be at the, uh, uh, uh, audience microphone sort of towards the middle back of the, um, of the room there.
[05:23.940 --> 05:27.180]  So, uh, uh, please welcome Sophie and Yan.
[05:47.720 --> 05:51.120]  So, uh, in case it's not obvious, that's Sophie.
[05:51.420 --> 05:52.100]  I'm Yan.
[05:52.460 --> 05:55.360]  There's been a little bit of confusion already.
[05:56.440 --> 05:58.860]  Uh, so hi, my name's Yan.
[05:59.240 --> 06:01.820]  Um, I am here to do the interview.
[06:02.280 --> 06:06.060]  I am, uh, currently the chief security officer at Brave.
[06:06.420 --> 06:13.300]  Previously, I worked at the Electronic Frontier Foundation on, uh, Let's Encrypt and other open source privacy software.
[06:14.260 --> 06:19.100]  So, um, if you don't mind, Sophie, I, I can say a few words about you based on what I know.
[06:19.640 --> 06:21.740]  So, Sophie is an amazing person.
[06:22.100 --> 06:24.360]  Uh, she's a former data scientist at Facebook.
[06:25.250 --> 06:40.420]  And on her last day of work in September 2020, she decided to publish an internal memo to Facebook employees, which was then leaked against her will and ended up making international headlines.
[06:40.840 --> 06:47.940]  So, I will read you a couple excerpts from that memo to give you the context for this Q&A session.
[06:48.640 --> 06:50.760]  Uh, excerpt one.
[06:50.760 --> 06:55.660]  Officially, I'm a low-level data scientist who's being fired today for poor performance.
[06:56.000 --> 07:12.420]  In practice, in the two and a half years I've spent at Facebook, I found multiple blatant attempts by foreign national governments to abuse our platform on vast scales, to mislead their own citizenry, and caused international news on multiple occasions.
[07:12.840 --> 07:24.120]  I have personally made decisions that affected national presidents without oversight and taken action to enforce against so many prominent politicians globally that I've lost count.
[07:24.920 --> 07:25.780]  Excerpt two.
[07:26.380 --> 07:34.180]  I consider myself to have been put in an impossible spot, caught between my loyalties to the company and my loyalties to the world as a whole.
[07:34.460 --> 07:46.000]  There was so much violating behavior worldwide that it was left to my personal assessment of which cases to further investigate, to file tasks, and to escalate for prioritization afterwards.
[07:46.400 --> 07:52.620]  With no oversight whatsoever, I was left in a situation where I was trusted with immense influence in my spare time.
[07:53.880 --> 07:54.760]  Excerpt three.
[07:55.000 --> 08:00.760]  I have made countless decisions in this vein, from Iraq to Indonesia, from Italy to El Salvador.
[08:01.300 --> 08:05.880]  Individually, the impact was likely small in each case, but the world is a vast place.
[08:06.800 --> 08:19.780]  Although I made the best decision I could based on the knowledge available at that time, ultimately I was the one who made the decision not to push more or prioritize further in each case, and I know that I have blood on my hands by now.
[08:20.860 --> 08:28.000]  So with that introduction, Sophie, could you tell our audience in your own words what was your role at Facebook?
[08:28.000 --> 08:29.420]  Yeah, absolutely.
[08:29.920 --> 08:31.220]  Is this playing?
[08:31.680 --> 08:32.180]  Can you hear me?
[08:33.400 --> 08:34.060]  It's good?
[08:34.260 --> 08:34.900]  Okay, great.
[08:35.460 --> 08:36.880]  So yeah, pleasure to talk with everyone.
[08:37.120 --> 08:37.220]  Thanks.
[08:37.340 --> 08:38.440]  This is my first hope.
[08:38.900 --> 08:42.640]  Anyways, so I joined Facebook in January of 2018.
[08:42.640 --> 08:44.700]  I was fired in September of 2020.
[08:44.780 --> 08:49.520]  And I was a data scientist, which means things are all over the place because of the new buzzword.
[08:49.740 --> 08:53.500]  At Facebook, it meant basically what's called a data analyst somewhere else.
[08:53.640 --> 08:55.240]  You look the numbers and tell people what they mean.
[08:55.580 --> 08:57.800]  And I was on the fake engagement team.
[08:58.740 --> 09:12.580]  By fake, I mean, for instance, actual fake accounts, hacked accounts, and which will probably not surprise many people in the room, but it's also a giant number of people who voluntarily give their accounts over to nefarious bad actors via social engineering to do bad things with them.
[09:13.060 --> 09:18.060]  So we call those self-compromised in that they're not quite fake, but they're inauthentic still.
[09:18.340 --> 09:20.560]  And by engagement, I mean likes, commentaries, etc.
[09:21.280 --> 09:30.160]  And so usually when I say this, what people immediately think of is Russian trolls, foreign interference, political activity, etc.
[09:30.820 --> 09:32.540]  But most people are not politicians.
[09:32.840 --> 09:36.040]  Most activity isn't apolitical on social media.
[09:36.360 --> 09:41.220]  The vast majority of this activity was not what you would consider very important.
[09:41.440 --> 09:54.260]  It was more in the spam category within the company organization in the sense of something that is vast in number, not too efficient, not too effective in individual cases.
[09:54.740 --> 09:59.500]  And so the work that you've heard about me doing was all work that I was doing in my spare time, essentially.
[10:00.980 --> 10:08.140]  It was, it was work that was close enough to my actual job, but would theoretically be left to much more important people to take care of.
[10:09.320 --> 10:15.660]  And, but, but in practice, those very much more important people were under-resourced, under-supported.
[10:16.080 --> 10:17.800]  They worked too hard with not enough pay.
[10:18.080 --> 10:22.200]  And so there ended up being a lot of things that they were not doing simply because they could not do them.
[10:23.020 --> 10:24.820]  I have the greatest respect for them.
[10:24.960 --> 10:27.620]  My disagreement is with the people who put them in those positions.
[10:28.000 --> 10:38.060]  And so I found myself increasingly working in the activity in the realm of sophisticated political activity, even though it wasn't quite my remit.
[10:38.280 --> 10:42.480]  The analogy here would be, say, a new policewoman hired in D.C.
[10:42.920 --> 10:49.120]  who is told to handle robbery and theft in D.C., by which the police chief means shoplifting, purse snatching.
[10:49.560 --> 10:50.320]  And she does that.
[10:50.500 --> 10:56.740]  But she also works in her spare time to track down a massive GRU ransomware operation attacks into D.C.
[10:56.840 --> 11:03.360]  hospitals to demand the residents steal their money, which is technically stealing things in D.C., but not exactly what she was expected to do.
[11:03.360 --> 11:06.060]  And so that's basically what I was doing.
[11:07.280 --> 11:11.980]  So in the course of your work as a data scientist, what did you find?
[11:13.060 --> 11:14.340]  Yeah, absolutely.
[11:15.020 --> 11:28.320]  So I eventually caught for certain two foreign national governments, red-handed, that were abusing Facebook on vast scales without even trying to hide it to manipulate their own citizenry.
[11:28.800 --> 11:37.280]  These were specifically the president of Honduras and the ruling political party of Azerbaijan.
[11:40.940 --> 11:50.940]  Honduras went through a deeply disputed election in 2017 that resulted in soldiers being sent on the streets to shoot civilian protesters after the police went on strike and refused.
[11:51.600 --> 12:02.520]  And this year, actually, the now-former Honduran president was extradited to here in New York City, in fact, where he's now on trial for smuggling drugs to the United States to finance his presidential campaign.
[12:02.760 --> 12:03.440]  Great person.
[12:05.220 --> 12:08.480]  And meanwhile, Azerbaijan is officially a democracy.
[12:08.520 --> 12:15.980]  It's so democratic that in 2013, they managed to accidentally release presidential election results a day before the actual election.
[12:18.000 --> 12:26.820]  And so, I mean, it's not too surprising that bad people would seek to abuse platforms in bad ways.
[12:27.380 --> 12:32.580]  What distinguished this was that, I mean, these were blatant violations of Facebook's rules in terms of service lists.
[12:32.740 --> 12:36.240]  And when I first saw this, I was very naive and silly.
[12:36.460 --> 12:37.920]  I thought, well, I found this.
[12:37.920 --> 12:39.480]  I'll just alert the authorities.
[12:39.600 --> 12:41.060]  I'll hand it over to take care of it.
[12:41.440 --> 12:42.000]  End of story.
[12:42.240 --> 12:45.100]  Which, I mean, I was never a security researcher.
[12:45.320 --> 12:45.840]  Maybe that shows.
[12:47.380 --> 12:56.760]  And so, in practice, for the start of a two-year Sisyphean or ideal, I mean, the way I would describe it is that everyone agreed it was terrible.
[12:57.000 --> 13:01.480]  No one agreed it was terrible enough to be their problem or their job or what to do about it.
[13:01.860 --> 13:07.040]  And so, in the end, it took almost a year for them to act in the case of Honduras after I caught them red-handed.
[13:07.040 --> 13:09.820]  It took more than a year to act in the case of Azerbaijan.
[13:10.160 --> 13:12.380]  In both cases, they came back right afterwards.
[13:12.600 --> 13:16.060]  The government of Azerbaijan is still happily active on social media today.
[13:16.540 --> 13:19.360]  But, I mean, Facebook is a company.
[13:19.580 --> 13:20.200]  They don't really care.
[13:21.140 --> 13:30.400]  And for people who aren't already familiar with the documents you leaked, what exactly were these governments doing that was unethical?
[13:31.720 --> 13:47.640]  So, essentially, these governments were using fake social media assets on vast scales to impersonate their own citizens and pretend to be their own citizens that supported themselves and opposed their opposition.
[13:48.020 --> 13:51.440]  I'm going to use Azerbaijan as an example.
[13:51.820 --> 13:55.360]  So, they had maybe 10,000 fake assets.
[13:55.360 --> 14:02.420]  My guess is maybe... I mean, I don't actually know how many people were involved because you can't pin fake Facebook accounts to these many people.
[14:02.520 --> 14:04.540]  But my guess is 1,000 or so big.
[14:04.600 --> 14:06.740]  They were creating comments on vast scales.
[14:06.980 --> 14:19.280]  And they were writing... Each month, they wrote 3 million individual unique comments all about how they were great, the opposition were terrible, and anyone who opposed them were Western spies or traitors or et cetera.
[14:19.780 --> 14:35.220]  And so, every time anyone posted about the opposition in Azerbaijan, any time someone discussed the opposition, any time a foreign news source such as BBC, Azerbaijan, Voice of America, Azerbaijan, Radio for Europe, Azerbaijan, any of them discussed the opposition,
[14:35.760 --> 14:43.880]  they would be immediately deluged by assaults and great comments about how they were traitors to the country and Azerbaijan did not need Western democracy and the things of that note.
[14:44.300 --> 14:47.220]  Which I'm sure must have had quite a chilling effect.
[14:48.240 --> 14:54.100]  And... I mean, people in this audience may be thinking, well, they should have just have created Facebook.
[14:54.460 --> 14:59.620]  But I mean, the thing to realize is that in countries like Azerbaijan, there's no independent media system.
[15:00.240 --> 15:06.340]  I mean, I think every independent media in Azerbaijan has been kicked from the country and it's basically foreign now.
[15:06.520 --> 15:09.200]  It's just in terms of being exiled, essentially.
[15:10.220 --> 15:19.560]  And the opposition really relies on social media for what is worse, because it's better than state-controlled media, which is even worse.
[15:20.200 --> 15:33.240]  Like, when I came forward with my findings in Azerbaijan, I was quite frankly a bit surprised at first by what the Azari political leader, Ali Karim Li, who was the main target of this harassment campaign, said.
[15:33.380 --> 15:39.780]  Because I would have expected him to say something very harsh on Facebook, but instead he said something very even-handed.
[15:39.940 --> 15:42.080]  He said something like, and I'm paraphrasing from memory.
[15:42.080 --> 15:53.340]  I mean, Facebook offers an important tool to the Azari opposition to get our message out and connect in a way that isn't affordable in government media.
[15:53.520 --> 15:59.120]  I mean, I thank Mark Zuckerberg for making Facebook, but he should hire someone who speaks Azari or something like that.
[15:59.220 --> 16:05.580]  He was very polite and frankly nice about it, because he could not afford to alienate Mark.
[16:05.580 --> 16:08.780]  He couldn't afford to leave Facebook, because the alternatives were even worse.
[16:09.320 --> 16:13.700]  And I think that's something that we and the rest of sometimes lose sight of.
[16:13.780 --> 16:18.780]  It's easy to say no one should use Facebook, but I mean, the alternatives can often be worse.
[16:20.280 --> 16:25.060]  So to go back a bit as to what these governments were doing, they were creating fake assets.
[16:25.300 --> 16:30.000]  And to most people in the audience, you might think of that as fake Facebook accounts, right?
[16:30.100 --> 16:33.040]  Like making accounts for users who don't actually exist.
[16:33.660 --> 16:43.300]  And I was kind of interested to learn that that's not what they were doing, because Facebook does sort of limit how many of these like fake user accounts you can create.
[16:43.500 --> 16:49.880]  So instead, they used a loophole where they made fake Facebook pages, which Facebook limits much less.
[16:50.100 --> 17:00.820]  And they were able to use those pages to create likes, posts, comments that look like they were from real users, but were actually from these like, you know, you call them inauthentic assets.
[17:01.160 --> 17:06.340]  So has Facebook done anything to solve that problem, or have they just left that loophole open?
[17:06.940 --> 17:09.200]  I mean, it was still active when I left the company.
[17:09.480 --> 17:11.660]  As far as I'm aware, it's still active today.
[17:12.300 --> 17:15.020]  I mean, still definitely being used by the government of Azerbaijan.
[17:15.320 --> 17:19.400]  I would call it a zero day exploit, except by now it's something like a 1500 day exploit.
[17:21.880 --> 17:27.000]  I mean, so just to back up a bit for people who do not use Facebook to try and explain what's going on.
[17:27.360 --> 17:33.920]  So I mean, when you use Facebook, most people have an account, but there's also something called a page, which is for public facing things.
[17:34.120 --> 17:40.720]  Like, for instance, HOPE might have a page on Facebook that is for the HOPE Hacker Organization conferencing.
[17:40.960 --> 17:43.380]  And they post, here's the update on the HOPE conference, etc.
[17:43.640 --> 17:46.080]  I mean, Joe Biden has a Facebook page.
[17:46.240 --> 17:49.040]  Donald Trump, well, has a Facebook page, but it's frozen.
[17:49.780 --> 17:55.720]  I mean, Narendra Modi, the Chinese government, the U.S. government, etc.
[17:55.720 --> 17:57.280]  All of these have Facebook pages.
[17:57.640 --> 18:00.500]  And that is the intent for Facebook pages.
[18:00.820 --> 18:04.420]  And so a single person can control a lot of Facebook pages.
[18:04.720 --> 18:14.840]  For instance, it might be that you're someone who works in the government and you control the U.S. government Facebook page, the White House Facebook page, and the Joe Biden Facebook page.
[18:14.900 --> 18:24.340]  I mean, this is because you're hired by them to run social media, because it's not actually Joe Biden who's typing in, I did so-and-so today, or et cetera.
[18:25.360 --> 18:32.180]  The thing is, there is no one checking that these pages are actual people.
[18:32.700 --> 18:43.620]  I mean, there are people who look that the pages aren't impersonating anyone important, but they can make a page that's called Joe Smith, and give it a name and a picture and pretend to be a person, and they can do things with it.
[18:43.780 --> 18:46.040]  And no one will stop me from doing it.
[18:47.660 --> 18:51.280]  I mean, I tried to make it a bit more difficult, but I got a lot of pushback.
[18:51.700 --> 19:00.000]  And Facebook has made it a lot easier for pages to do things, because the pages growth team is like, yay, that increases our numbers and makes people use pages more.
[19:00.240 --> 19:03.560]  And so here's a fun fact that I was shocked when I discovered it.
[19:04.440 --> 19:16.820]  Of Facebook activity by pages, globally, worldwide, 3% of activity from pages on posts by other pages, were from this Larry network, by the Larry government, worldwide.
[19:17.580 --> 19:33.800]  And so somewhere in Facebook, I'm sure there's a team that's very proud of their numbers, and very surprised when it went temporarily down in late October, November 2020, when I got this taken down briefly, post my departure, and they came right back.
[19:34.320 --> 19:43.360]  But I mean, I mean, ultimately, that's, that's what happens when you have an organization that focuses on numbers, rather than what seems actually important.
[19:44.400 --> 19:44.800]  Okay.
[19:45.420 --> 19:54.160]  So in your work as a data scientist, what did you look at to determine that these were, how did you find out this loophole was being exploited, basically?
[19:54.880 --> 19:57.400]  In this specific case, it was very silly.
[19:58.400 --> 20:10.200]  So I, so, so at the start, I was, so at the start, I was just looking into fake likes, et cetera, in specific things that were not too important, too important.
[20:10.640 --> 20:15.680]  And there was, and this was a very new team, I think it was the second or third person to join.
[20:16.180 --> 20:18.920]  And right, right, right when I was hired.
[20:19.900 --> 20:24.620]  And, and so I took a while to get on the ground, I started looking into how this was being used in politics.
[20:25.520 --> 20:31.120]  And so, and so I looked for intersect, we, we had a database of activity we believed was fake.
[20:31.360 --> 20:35.980]  We also had a database of activity we think is political, because the political team does that.
[20:36.120 --> 20:40.100]  And so I think, okay, our union, these two databases, and see what I got, I'm curious.
[20:40.400 --> 20:47.480]  And so I got activity, I got activity in India, I got activity in Indonesia, really, in what we would call the global source.
[20:48.380 --> 20:50.460]  And I was putting together a report for leadership.
[20:51.120 --> 20:54.720]  And on the rest was one of these cases, actually, the underlying president was the only one involved.
[20:54.900 --> 21:00.260]  And I was putting together a presentation for leadership, to tell them, this is bad, and we should do something about it.
[21:00.440 --> 21:08.040]  And then I was, I was just opening the underlying president's page, and clicking on who liked him, so I could get a screenshot for a PowerPoint.
[21:08.240 --> 21:08.920]  That's all I was doing.
[21:08.980 --> 21:12.980]  And I saw Wade, this is weird, because all these people who are liking him, aren't people, they're pagers.
[21:13.300 --> 21:14.320]  This is really silly.
[21:15.000 --> 21:20.760]  It would be like, it would be like, I don't know, you walk into a normal person conference, and everyone has pink hair.
[21:22.140 --> 21:24.380]  I don't know, I'm not great at my analogies.
[21:27.880 --> 21:30.720]  Yes, I mean, I said, a normal person conference.
[21:32.400 --> 21:34.220]  Sorry, apologies to everyone here.
[21:35.120 --> 21:38.960]  You know, like a fancy corporate conference, and everyone has pink hair, that would be a bit weird.
[21:40.280 --> 21:43.300]  And so, and so anyways, I looked more into that.
[21:43.420 --> 21:45.920]  And I was like, wait, is no one paying attention to this?
[21:46.080 --> 21:47.760]  Well, I guess it's a zero-day exploit.
[21:47.780 --> 21:49.540]  And I was like, okay, this is very obvious.
[21:49.700 --> 21:49.960]  We're close.
[21:50.160 --> 21:50.760]  It's a loophole.
[21:50.940 --> 21:52.360]  It was not that obvious.
[21:52.840 --> 21:54.440]  I was very naive and silly.
[21:55.860 --> 22:00.560]  And, and, but I mean, that, that was a dramatic, that was a dramatic and silly case.
[22:00.680 --> 22:09.440]  It was easy from the start, because like, they didn't even bother to hide, because usually in like hacking, in hacking, like you want, it's hard to attribute attribution.
[22:09.680 --> 22:11.100]  You don't know who's responsible.
[22:11.420 --> 22:19.860]  In this case, it was the equivalent of some, if someone hacked into a database and lined, and, and left the file signed with a private, signed with a private key saying, I did it or something.
[22:20.140 --> 22:34.180]  So, so you're saying that these Facebook pages, which were liking the, um, Azerbaijan or Honduras, you know, president's posts were, are, were they just literally being run by real Facebook accounts that belonged to politicians in the government?
[22:34.180 --> 22:44.160]  In this case, there were 200 fake Facebook pages that was run literally by the page administrator of the Honduran president who was given special access to like posts on his behalf.
[22:44.340 --> 22:46.020]  I mean, like usually you have to be careful.
[22:46.200 --> 22:52.700]  This was, they didn't bother to be careful because they were so blatant and, blatant and didn't care and weren't afraid.
[22:53.080 --> 22:55.920]  Like, I'm not, I'm not a super genius.
[22:56.160 --> 22:57.860]  I had no super, I had no special training.
[22:58.340 --> 23:00.600]  Like, I had no special training.
[23:00.760 --> 23:01.500]  I'm not a super genius.
[23:01.500 --> 23:02.840]  I've met super geniuses.
[23:02.980 --> 23:03.840]  I'm not one of them.
[23:03.940 --> 23:06.220]  And I still caught two national governments right-handed.
[23:06.340 --> 23:09.720]  And that's not because I had anything special.
[23:09.920 --> 23:14.100]  It's because no one had bothered to look for them before because anyone who could have, could have found this out.
[23:14.880 --> 23:17.940]  So you said that Facebook has not yet closed this loophole.
[23:18.140 --> 23:25.360]  So what was, I mean, I assume that means their response to you finding this was just, you know, shrug or like, what was, what did they do?
[23:25.580 --> 23:27.100]  What happened after you told them?
[23:27.240 --> 23:31.360]  I mean, the way we describe it is that everyone agreed that it was terrible.
[23:31.360 --> 23:36.820]  And no one agreed on what the correct solution was, whose job, and more importantly, whose job it was.
[23:37.420 --> 23:38.740]  I'm going to use an analogy.
[23:39.540 --> 23:40.840]  So, I mean, we're in the U.S.
[23:41.060 --> 23:43.660]  There are, unfortunately, a lot of mass shootings in the U.S.
[23:44.440 --> 23:49.840]  After each mass shooting, there's absolutely no one who says, mass shootings are great, mass shootings are okay.
[23:50.400 --> 23:58.380]  What happens instead is that everyone agrees that mass shootings are terrible, but people don't agree on whose job it is to fix it, what the correct solution is.
[23:58.460 --> 24:01.460]  Some people say we need more guns in armed societies, a polite society.
[24:01.720 --> 24:03.760]  Some people say we need fewer guns, we need gun control.
[24:03.860 --> 24:05.840]  Some people say the correct answer is thoughts and prayers.
[24:06.120 --> 24:08.460]  And so, that's the only thing that happens in the end.
[24:09.560 --> 24:14.640]  And so, like, I talked, because, I mean, I don't blame the individual people.
[24:14.840 --> 24:25.820]  Like, if someone, if, if you are at, if you work in a white hat office space, and someone who's not on your team, who's lower level than you, goes up and says, hey, you should do this extra work.
[24:25.960 --> 24:27.700]  You're not going to be like, oh, great, I should do that.
[24:28.940 --> 24:34.780]  But, I mean, so ultimately, I blame the leadership who should have found someone to put this aside.
[24:36.720 --> 24:40.280]  But, yeah, like, I tried to get, I tried to get pages to own it.
[24:40.720 --> 24:42.120]  They didn't think it was a problem.
[24:42.240 --> 24:43.180]  I tried to get newsfeed.
[24:43.300 --> 24:53.380]  I tried to get, I tried to get social act of violence, strategic response, civic integrity, all these names that mean absolutely nothing to these people in this room, but our different Facebook teams.
[24:53.820 --> 24:56.520]  And all of them agreed it was terrible and not their problem.
[24:57.240 --> 25:14.380]  And so, eventually, I found out that the best solution was not to, was not to go through the proper channels, but to make noise internally, because, I mean, so, so Facebook has Facebook, the product, and it also has something called workplace, which is basically Facebook inside the company.
[25:14.760 --> 25:18.260]  And any employee could make posts inside workplace and hope it goes viral.
[25:18.600 --> 25:19.560]  And so, that's what I did.
[25:19.800 --> 25:23.240]  And eventually, one of them got viral, and they got people to be like, wait, what the f*ck are we doing?
[25:23.740 --> 25:27.940]  And then, and that caused enough pressure for people to act a tiny bit.
[25:28.620 --> 25:31.980]  And so, that was my great, brilliant method to getting change.
[25:32.160 --> 25:33.740]  I mean, I also got pushback.
[25:33.940 --> 25:39.520]  Like, I got, I got pushed, I got criticized by a company vice president a few times, for instance.
[25:39.960 --> 25:42.500]  Like, I'm super low, I'm super, I was super low level.
[25:42.660 --> 25:45.100]  I was literally one level above a new hire straight out of college.
[25:45.160 --> 25:48.360]  So, I shouldn't have been interacting with him in the first place.
[25:48.600 --> 25:51.640]  But my life got very weird around that point.
[25:52.500 --> 25:56.720]  So, what about the consequences of you coming out publicly with this?
[25:56.820 --> 26:00.880]  Has this put any PR pressure on Facebook to start fixing these issues?
[26:01.080 --> 26:03.880]  You know, perhaps hiring more people to work on these problems?
[26:04.120 --> 26:08.460]  Like, literally, to me, it sounds like there was just you looking at this in your spare time.
[26:08.600 --> 26:17.660]  When, you know, a company of that scale should have multiple employees, perhaps even teams, investigating this kind of activity, especially in political situations.
[26:17.880 --> 26:22.920]  So, I do want you to take a step back and address the larger question of who works on this, and why didn't they work on this?
[26:23.280 --> 26:27.600]  So, there's a team that's called Shred Intelligence that, in theory, should work on this.
[26:27.740 --> 26:30.600]  They put out monthly coordinated inauthentic behavior reports.
[26:30.920 --> 26:35.560]  Like, you might have heard of them putting out reports when they took about Russian interference, or et cetera.
[26:36.180 --> 26:42.980]  Or about, for instance, the TPUSA associate Raleigh Forge was writing fake accounts in mid-2020.
[26:43.140 --> 26:44.960]  You might have heard of that when it was taken down.
[26:46.180 --> 26:49.440]  And they're a very insular team, very secretive.
[26:50.060 --> 26:52.680]  Many of them might think former CIA's books, or et cetera.
[26:52.940 --> 26:56.140]  They're highly trained, and they're also very few and work way too hard.
[26:56.880 --> 27:05.460]  And the general organization contains a lot of similarly important people, like, for instance, espionage, which alerts people, hey, there's a foreign spy who's looking at your Facebook profile.
[27:05.540 --> 27:06.440]  You might want to know that.
[27:06.440 --> 27:10.240]  And so, in theory, that should be their job.
[27:10.740 --> 27:19.300]  In practice, in my experience, they tend to focus reactively on what gets attention.
[27:21.160 --> 27:26.420]  In the sense, because usually Facebook responds to cases when they come from outside the company.
[27:26.760 --> 27:35.100]  When there's a foreign NGO, when there's a political party that says, hey, this looks weird, when the media reports something really strange is going on, then Facebook looks into it.
[27:35.100 --> 27:44.620]  Because there's people outside the company who care and can hold Facebook responsible, and that can cause Facebook to get bad news reports, and that will affect their ability to make money.
[27:45.560 --> 27:55.020]  In contrast, I was an employee who went out and found this on my own, completely proactively, without anyone outside the company to hold them responsible.
[27:55.620 --> 27:57.620]  In theory, my loyalty was to the company.
[27:58.100 --> 28:10.240]  I mean, I kept trying to use arguments like, well, this is so obvious, eventually someone will notice, and if they do, I mean, Facebook has so many leaks, if it ever gets all that we said on this for a year, it will look terrible.
[28:10.400 --> 28:11.440]  We got killed in the press.
[28:11.700 --> 28:16.320]  And this became a self-fulfilling prophecy, because I was the one who leaked it, but we didn't know that at the time.
[28:17.500 --> 28:22.580]  But, I mean, the point I'm trying to make, though, is that, like, ultimately, Facebook is a company.
[28:22.580 --> 28:34.000]  We don't expect Philip Morris to go out and have a special division that reimburses Medicare whenever someone gets lung cancer, or have a special division that tries to make cigarettes less addictive.
[28:34.320 --> 28:37.920]  It would frankly not be very effective if that were the case.
[28:38.080 --> 28:44.680]  They would say, okay, that's nicotine in cigarettes, that's etc., and people would be like, no, that would cost a profit margin, or whatever.
[28:45.300 --> 28:58.480]  And so, like, and so, and so what I'm trying to say, ultimately, is that there were people who were working on this, but because of Facebook's prioritizations, they were not working on the specific cases they found.
[28:58.740 --> 29:01.220]  They were not considered important to Facebook.
[29:01.380 --> 29:06.520]  I was actually told at one point that this work was not considered important to Facebook.
[29:07.400 --> 29:13.360]  And if it was, I should just prove it by doing nothing, and that didn't become a PR fire, so that people would care about this.
[29:14.160 --> 29:26.760]  And this sort of logic makes sort of sense for other areas, such as, for instance, misinformation, hate speech, etc., which are content-based issues in which the average person can pay attention and come to some conclusion.
[29:27.500 --> 29:32.320]  If someone on the Internet says, the moon is made out of cheese, that's not a huge deal.
[29:32.920 --> 29:40.660]  If a politician says the moon is made out of cheese, and that starts a giant Internet-filled memes with eat the moon.
[29:40.780 --> 29:41.340]  The moon is cheese.
[29:41.500 --> 29:43.500]  The moon is so delicious, eat it like a tide bald.
[29:44.080 --> 29:49.400]  And suddenly a giant crowdstorms Cape Canaveral trying to hijack a rocket up to the moon to eat the cheese.
[29:50.140 --> 29:53.380]  I mean, I'm using the absurd example intentionally on purpose.
[29:53.940 --> 29:58.060]  I mean, that would be very important and very bad, and it would get a lot of attention.
[29:58.520 --> 29:59.800]  And the first would be not...
[29:59.800 --> 30:02.600]  some random person doing it would not get much attention.
[30:03.100 --> 30:05.320]  And so, in this case, it would be a decent proxy.
[30:05.960 --> 30:09.160]  But the problem with fake accounts is that the goal is to not be seen.
[30:09.660 --> 30:11.940]  I mean, I mean, an analogy you're probably familiar with.
[30:12.060 --> 30:16.500]  If everyone knows of a hacker, they're probably a bad hacker who's not good at being hidden.
[30:17.260 --> 30:18.220]  I mean, sorry.
[30:18.760 --> 30:21.020]  I mean, and that's if... well, not necessarily.
[30:21.360 --> 30:22.440]  I mean, it depends on the type of hacker.
[30:22.740 --> 30:23.660]  Like, white hat is okay.
[30:23.800 --> 30:26.020]  I'm talking more about the really black hat types.
[30:26.420 --> 30:29.520]  Well, like, if everyone knows that someone is a spy, they're a terrible spy.
[30:29.760 --> 30:31.080]  Unless his name is James Bond.
[30:31.180 --> 30:32.500]  I don't know how James Bond does it.
[30:35.280 --> 30:35.800]  Yeah.
[30:36.300 --> 30:44.400]  And so, the point I'm trying to make is that, fundamentally, the attention is misallocated for fake accounts and that sort of thing.
[30:44.780 --> 30:46.740]  People pay attention to what gets attention.
[30:46.980 --> 30:49.200]  That's not necessarily what's effective.
[30:49.460 --> 30:51.280]  It's what's bad enough to be caught.
[30:52.420 --> 30:55.900]  And so, they allocate it to the wrong areas.
[30:56.220 --> 30:59.500]  And so, Facebook pays attention to the wrong things.
[31:01.000 --> 31:04.740]  And so, I was providing context, and there was an actual question.
[31:04.940 --> 31:05.800]  And I forgot your question.
[31:05.900 --> 31:06.360]  I'm sorry.
[31:06.820 --> 31:09.160]  Yeah, let me try to remember it for a moment.
[31:09.260 --> 31:11.160]  I swear I didn't try to do that on purpose.
[31:12.940 --> 31:14.200]  What was my question?
[31:15.380 --> 31:26.200]  Well, actually, I have a somewhat related question, which is, you know, there's some things which are easy for third parties, like people in the audience, to detect, and some things that are not.
[31:26.480 --> 31:35.360]  And really, the only people who can find it are people who work internally at Facebook, which from what you've told us sounds like is currently nobody because you were fired.
[31:35.620 --> 31:37.400]  You were the only person doing it before.
[31:38.480 --> 31:46.320]  It's things like looking at, you know, comments or likes on a politician's Facebook post and seeing if those are coming from pages.
[31:46.640 --> 31:50.020]  If I'm not mistaken, it's something that, like, anyone can do, right?
[31:50.020 --> 31:51.560]  It's just Facebook has rate limits.
[31:51.820 --> 31:53.860]  They have probably, like, anti-scraping measures.
[31:53.860 --> 31:57.220]  Yeah, I put up the rate limits myself, so we're not very effective.
[31:57.420 --> 32:00.040]  I tried to make them narrower, but it didn't.
[32:00.240 --> 32:01.420]  But there was pushback.
[32:01.580 --> 32:03.520]  People were like, it would affect our performance.
[32:03.880 --> 32:04.280]  Okay.
[32:04.440 --> 32:23.360]  So I guess my question is, is there anything people who don't work at Facebook can do to help detect this kind of activity, to bring attention to the fact that this is happening, and put more external pressure on Facebook to close these loopholes, dedicate more internal resources to fixing them,
[32:23.480 --> 32:24.280]  and so forth?
[32:24.420 --> 32:24.580]  Yeah.
[32:25.000 --> 32:27.340]  So I'm going to decompress that into several parts.
[32:27.540 --> 32:29.880]  First is, can people outside detect these?
[32:30.320 --> 32:47.540]  And the answer is, frankly, it's really hard in the same way that, that it's, that, I mean, because, at least, so I'm not going to go into exactly how I called fake accounts into governments, because, for instance, those people will also listen to that talk,
[32:47.600 --> 32:51.360]  and they will tell them, I called them to X, Y, Z, they will know not to do X, Y, Z in the future.
[32:52.840 --> 32:56.360]  But, but very broadly, I used metadata on large levels.
[32:56.560 --> 33:10.520]  I know it's unpopular, but, I mean, but ultimately, it's, to actually associate this account with that person, this, this user is based in this area, I mean, you have to use metadata on some level, if you want to do any sort of attribution.
[33:11.680 --> 33:26.420]  And, there are things that you can look at, as, as an external researcher, but none, but most of them tend to be imperfect, because, in the same way that, if, if people in this audience go out and start playing, it's someone in, are there any spies in this audience?
[33:26.560 --> 33:27.240]  Let's do a survey.
[33:27.820 --> 33:33.000]  I mean, this will not be very effective, because the goal is to, well, those people is to hide, and pretend they're not a spy.
[33:33.140 --> 33:35.640]  And so, you see, this person's behaving very unusually.
[33:36.160 --> 33:40.080]  Actually, they're just sleep deprived, because they spent last night doing hacker karaoke.
[33:41.840 --> 33:44.160]  This person is dressed up as a spy.
[33:44.480 --> 33:47.380]  Actually, they're about on their way to a hacker dress-up party.
[33:48.980 --> 33:50.000]  I don't know.
[33:51.080 --> 34:02.520]  And so, it's, it's, it's, it's, it's, it's almost impossible to reliably distinguish the fake accounts from the real people, that, that, without internal data.
[34:03.140 --> 34:09.060]  And that's a lot of the issue at hand here, that only Facebook has its data, and we, and, and they're not going to give it up.
[34:09.940 --> 34:16.760]  So, in terms of actual paths forward, I mean, I've tried to put public pressure on Facebook around.
[34:16.980 --> 34:19.400]  I have to say, by now, it's not very successful.
[34:19.820 --> 34:24.520]  I'm aware of only one change Facebook has made since I left that might be related to me.
[34:24.520 --> 34:30.000]  That change is that they have cracked down on internal conversation that isn't directly related to, to work.
[34:30.260 --> 34:45.140]  Which is certainly a change, but not one that I would necessarily recommend, or, I mean, it's not too different from, say, what Bader Routes did to, it's, to the, to, to the mass protests with Tsikonoskaya coming forward and all that.
[34:46.000 --> 35:04.160]  And so, and, and, and so, like, I've testified to a number of governments, and, and, and, and, and what they, and what they advised to try and, and to plug that gap is to essentially do sort of red team style penetration testing.
[35:04.560 --> 35:08.340]  That is, because, because right now, you, you don't even have a good idea.
[35:08.500 --> 35:10.440]  It's, how does Facebook do with bots?
[35:10.620 --> 35:11.640]  How does Twitter do with bots?
[35:11.780 --> 35:16.100]  I mean, you know, so, but so little about that, we can have giant arguments about it on Twitter with Elon Musk.
[35:17.520 --> 35:23.980]  And, like, no one has the knowledge right now besides the social media companies who are not going to give that up.
[35:24.060 --> 35:36.860]  And so, the only effective way I can see to do it is to do red team style penetration tests with ideally governments or trusted white tech organizations that would do, would do red penetration tests.
[35:37.000 --> 35:42.680]  You set up, you set up 10 fake, 10 networks of fake accounts on Reddit social, Reddit, Twitter, Facebook.
[35:42.880 --> 35:44.900]  And then you say, Facebook caught one out of 10.
[35:45.100 --> 35:45.780]  Reddit caught zero.
[35:45.940 --> 35:46.540]  Twitter caught zero.
[35:46.720 --> 35:47.320]  They're all terrible.
[35:47.440 --> 35:48.400]  Facebook is the least terrible.
[35:48.700 --> 35:49.900]  These are made up numbers, of course.
[35:50.180 --> 35:54.140]  And you could do the same thing with any, any other problem on social media in theory.
[35:54.820 --> 36:06.880]  For instance, if people, you were worried that my, my political philosophy is getting censored incorrectly, then you can make posts that are perfectly fine and report them and see how many of them are incorrectly taken down.
[36:07.220 --> 36:09.120]  Same with other political philosophies.
[36:09.220 --> 36:13.100]  You think Facebook is not cracking down on hate speech that violates its terms of service.
[36:13.580 --> 36:17.300]  You make hate speech, you report it, you don't report it, you see how much of it gets taken down.
[36:17.720 --> 36:28.160]  I mean, same process applies for, for, for terror, for terrorism, child pornography, anything though, some of these would be very hard to do in a non-crimeful way.
[36:28.620 --> 36:35.080]  And because right now, the fundamental problem is that you can't solve a problem that you don't understand.
[36:35.440 --> 36:40.000]  And we, and because of this information gap, we don't really understand what's going on on social media.
[36:41.340 --> 36:50.580]  You brought up a really interesting point earlier, which is, I assume we're in an audience of people who are generally pretty hateful towards Mark Zuckerberg's products.
[36:51.080 --> 36:57.220]  And I know we will eventually get to a point in this talk where people will ask, why do we still use Facebook?
[37:06.140 --> 37:16.120]  Why there's really no alternative to Facebook and for people to express their opinions openly as an alternative to state controlled media.
[37:16.880 --> 37:24.620]  So, but it seems that if we have an alternative, like Mastodon, there will still be this problem, right?
[37:24.720 --> 37:31.040]  Of like fake accounts and, and the service providers needing to detect that activity and figure that out.
[37:31.160 --> 37:40.120]  So how do you kind of square this need for moderation and control by the platform versus like our, you know, love of decentralization?
[37:41.680 --> 37:48.300]  Like, I don't, it's, I want to be clear, I am not an expert on Mastodon, decentralized social media, etc.
[37:48.820 --> 38:03.780]  My general vague impression is, it's from decentralized systems, and for instance, decentralized financial systems is that you, it's that it's very hard for the centralized system to crack down on very broadly fraud and bad actors.
[38:04.060 --> 38:07.980]  You need some, and I think crypto is discovering that right now, quite frankly.
[38:09.040 --> 38:11.620]  I hope that's not too controversial for me to say.
[38:12.500 --> 38:25.120]  And like, it, but, but it's definitely true that finances are much more interesting to bad actors and fraud than say, a tiny, a tiny Mastodon server.
[38:25.400 --> 38:44.520]  I do think that Mastodon will, might have growing pains if people adopt it en masse, but I don't, and, but, but going, but go, and going back, you were talking, you were talking about centralized regulation, well, not regulation, just moderation, really.
[38:44.920 --> 38:54.740]  I do want to break moderation into separate, into, into identity-based categories and content-based categories, though, because I think it's important and often gets confused.
[38:54.740 --> 38:58.000]  People say fake accounts, that's like misinformation, right?
[38:58.620 --> 39:00.580]  They're, they're actually completely separate problems.
[39:01.500 --> 39:02.940]  Misinformation, hate speech, et cetera.
[39:03.120 --> 39:05.440]  This is the, this, it's a content-based problem.
[39:06.000 --> 39:06.980]  It doesn't matter who you are.
[39:07.160 --> 39:12.380]  If you, if you say rabbits are the same species as cats, this is a lie.
[39:12.580 --> 39:13.340]  It's misinformation.
[39:13.800 --> 39:15.060]  You could be a pet store owner.
[39:15.180 --> 39:16.340]  You could be a 10-year-old kid.
[39:16.780 --> 39:19.420]  It's, you could be, uh, a politician.
[39:19.660 --> 39:20.840]  They're different species.
[39:22.440 --> 39:24.160]  I like how you're laughing, yeah.
[39:24.620 --> 39:25.120]  That's okay.
[39:25.280 --> 39:26.600]  I will never say that.
[39:26.980 --> 39:27.460]  Exactly.
[39:28.100 --> 39:29.540]  Sorry, I hope you're not offended.
[39:29.740 --> 39:30.720]  I know you love funnies.
[39:31.320 --> 39:32.920]  I hope Zuki's doing okay.
[39:33.380 --> 39:44.600]  Anyways, um, and if you, if you say, if you set up 10,000 fake accounts on social media to say rabbits are different species from cats, there's nothing wrong with this statement.
[39:45.120 --> 39:47.940]  Except for the fact that I'm using fake accounts to say it.
[39:48.080 --> 39:49.980]  And then this becomes a new trending thing on Twitter.
[39:50.280 --> 39:52.360]  And I'm drawing out the voices of actual people.
[39:53.540 --> 40:01.540]  And, and I can yell afterwards, okay, Facebook, Twitter is taking down my correct factual statement and that cats and rabbits are different.
[40:01.760 --> 40:03.180]  But this has nothing to do with the content.
[40:03.400 --> 40:07.720]  It has everything to do with the fact that I'm pretending to be a giant crowd when I'm a single person.
[40:09.160 --> 40:23.660]  And, and, and, and so, I mean, content moderation is, it's deeply controversial because, I mean, because people disagree on where to draw the line and how much control to give people over what we can talk about.
[40:23.840 --> 40:30.080]  And, I mean, and, and I think that reasonable people can have different conclusions over where to draw the line.
[40:30.480 --> 40:48.780]  But there's, I have personally seen very few people disagreeing with the idea that, with the idea that, that, that, that fake accounts taking over the ecosystem should be taken down, that, that governments should be allowed to create tens of thousands of fake accounts to distort voices and pretend to be real people.
[40:49.680 --> 40:56.560]  And, and it, because this draws out the voices of the regular people, just like allowing ballot boxes to be stuffed with fake votes.
[40:57.180 --> 41:00.940]  And it, I mean, it's important to stop that to protect the sanctity of the ballot.
[41:01.160 --> 41:04.620]  It isn't some, it isn't stopping the right to vote to remove the fake ballots.
[41:04.820 --> 41:07.420]  It's actually necessary to, to, to do it.
[41:07.900 --> 41:10.360]  And I hope this is making sense as an analogy.
[41:11.540 --> 41:18.880]  And so, and, and so, with regard to content moderation, because I think you were heading in that direction.
[41:20.040 --> 41:27.320]  I, I mean, this, there has been a lot of debate over censorship versus protecting people lately.
[41:28.100 --> 41:32.460]  My personal belief, I want, I do want to really make clear this personal belief.
[41:32.640 --> 41:35.780]  I did not personally work on distribution, content moderation, etc.
[41:36.140 --> 41:42.720]  But my personal belief is that this is sort of a distraction from the actual issue, which isn't that bad content is being posted.
[41:43.060 --> 41:49.500]  It's that bad content, it's been, it's that misinformation conspiracy theories are being distributed when they would not have been previously.
[41:50.040 --> 41:52.100]  I mean, there were always conspiracy theories.
[41:52.540 --> 41:59.960]  Like 60 years ago, here in the United States, there was a conspiracy theory that fluoride was the communist conspiracy to sap your bodily fluids.
[42:00.980 --> 42:03.520]  But no, I mean, no one would give them the time of the day.
[42:03.760 --> 42:08.080]  I mean, I think it showed up in the John Birch Society a bit, but I mean, it did not show up on CNN.
[42:08.220 --> 42:10.060]  It did not, was CNN around then?
[42:10.200 --> 42:10.500]  Sorry.
[42:11.080 --> 42:12.540]  Well, that's why it didn't show up on CNN.
[42:13.040 --> 42:14.500]  It didn't show up in the New York Times.
[42:14.580 --> 42:16.520]  It didn't show up in the Washington Post.
[42:16.620 --> 42:19.340]  And if it did, it would only be to ridicule it, really.
[42:20.220 --> 42:28.600]  And because essentially, back 60 years ago, so large media organizations served as gatekeepers.
[42:29.420 --> 42:37.340]  If your message wasn't deemed fit for publication in the MIT, it would be much harder to get an audience.
[42:38.140 --> 42:40.660]  And social media has disrupted all of that.
[42:40.820 --> 42:46.300]  And I was very optimistic about this when it started happening like 10, 20 years ago.
[42:46.420 --> 42:48.060]  And it has had a lot of positive impacts.
[42:48.280 --> 42:53.360]  Like, for instance, I mean, it's helped... it's helped... I mean, counter-cultural people find each other.
[42:53.500 --> 42:55.020]  It's helped the queer community find each other.
[42:55.180 --> 43:02.580]  But it's also true that, I mean, with the end of gatekeepers, I mean, anytime you make changes, there's going to be consequences.
[43:02.720 --> 43:04.740]  Some of them that aren't positive.
[43:05.020 --> 43:09.680]  And I think it's all well and good to talk about, let's get rid of the gatekeepers.
[43:09.880 --> 43:11.160]  Everyone can do whatever.
[43:11.160 --> 43:14.540]  And... but I think we're still grappling with the consequences of that today.
[43:15.620 --> 43:20.820]  Like, I was very optimistic about getting rid of gatekeepers when it started, but back 20 years ago.
[43:21.040 --> 43:27.340]  And now I'm wondering if it would be worth it to try and stuff the genie back in the bottle, quite frankly.
[43:29.460 --> 43:48.440]  I had a thought, I think, somewhere around 2016 or so, and especially that compounded with the rise of technology like GPT-3, where it became very easy for people to make a lot of fake content that looked real without having to actually do a lot of manual effort.
[43:49.280 --> 43:56.620]  That maybe the solution is for people just to never trust anything they see online, except from a very small number of sources.
[43:56.920 --> 44:11.840]  For instance, I assume the reason governments buy fake, or not buy, but generate these fake likes and posts and re-shares and so forth is to make themselves look more popular to their populace.
[44:12.000 --> 44:25.400]  But if on the other side of that, the people looking at that just didn't believe it, you know, if everyone just looked at a post and said, all those likes could be fake, all those comments could be fake, I'm going to disregard that because it's not from a trusted source,
[44:25.580 --> 44:28.180]  then perhaps we wouldn't have this problem to begin with.
[44:28.340 --> 44:34.240]  Do you think we're headed in that direction where everything online is just becoming fundamentally untrustworthy?
[44:35.140 --> 44:37.760]  I think we are headed somewhat in that direction.
[44:38.080 --> 44:43.500]  And I don't think, and I wouldn't call it a positive change, because society does rely on trust to some degree.
[44:43.840 --> 44:46.400]  I mean, I can trust that they can go on stage and no one mobs me.
[44:46.580 --> 44:51.920]  I can trust that they go to a Starbucks and buy a sandwich and the waitress isn't going to poison my food.
[44:52.080 --> 44:53.140]  She's not going to steal my money.
[44:53.660 --> 44:59.840]  I can trust that I'm spent using my credit card and people don't have a skimmer set up to steal it.
[45:01.140 --> 45:08.480]  And if you distrust everything, then it loses a lot of society.
[45:08.620 --> 45:12.640]  It loses the ability to make friends online to talk to one another.
[45:12.840 --> 45:15.820]  It loses a lot of what made it vulnerable in the first place.
[45:16.120 --> 45:23.860]  And not only that, when nothing is trusted, what that really benefits is the people who are untrustworthy to begin with.
[45:25.600 --> 45:32.920]  A society in which everyone is assumed to be a criminal, the only people who haven't lost anything are the people who were already criminals.
[45:33.520 --> 45:43.580]  A society in which no one is trusted, the people who benefit the most are the people who are the least trustworthy, because now you're dragging everyone down to their level.
[45:45.040 --> 45:58.360]  A lot of why authoritarian governments feel the need to do this sort of thing in the first place is that in dictatorships, the appearance of popularity can be more important than actual popularity.
[45:58.820 --> 46:07.060]  Because, I mean, if you're going to be arrested for saying you oppose Putin, and if you're going to be arrested for saying you oppose the war, then you're not going to say that you oppose the war.
[46:07.260 --> 46:14.980]  But how do you overthrow the government if everyone else is saying, oh yeah, I support the government regardless of what they actually think?
[46:15.180 --> 46:25.620]  It's really hard to go and find other people who think similarly, to find people who... to know that you're actually in the majority, to know that maybe that everyone believes like you do.
[46:25.840 --> 46:30.240]  Everyone is being like, I hate the government, but I'm going to go along with it because everyone else does.
[46:30.320 --> 46:33.180]  And they're all thinking the same thing because they don't know they all hate the government.
[46:34.400 --> 46:51.040]  And, I mean, with the communist regimes in the Eastern Bloc, like for instance, a lot of the reason the governments felt the need to bust in protesters, sorry, to bust in supporters en masse in rallies, was that they needed an appearance of support.
[46:51.260 --> 46:54.340]  If they held a rally and no one showed up, that would be a sign of weakness.
[46:55.500 --> 47:00.700]  But that had its risks, like the fall of Nicolae Ceaușescu in Romania, that happened.
[47:00.860 --> 47:05.940]  He bust in a giant crowd of 100,000 people, and that crowd turned on him in the middle of the speech.
[47:06.420 --> 47:07.740]  And that started a revolution.
[47:08.120 --> 47:11.480]  The army turned on him the next day.
[47:12.540 --> 47:15.260]  Within a few more days, he was given a show trial and executed.
[47:16.240 --> 47:26.060]  But that requires people to trust that when people are opposing the dictator, they are actual people and not the opposition making that these are my people who agree with me.
[47:26.180 --> 47:29.740]  Like if no one trusts anyone, like that's also the dictator's job accomplished.
[47:29.960 --> 47:31.960]  No one will trust anyone enough to rise up.
[47:32.680 --> 47:35.260]  And I hope this analogy is making sort of sense.
[47:37.360 --> 47:45.160]  So I think I want to step back a bit and talk about your experience leading up to this this act of whistleblowing.
[47:45.680 --> 47:52.960]  So I know many people here are critical of people who make the decision to work at Facebook in the first place.
[47:53.620 --> 47:56.020]  What brought you to work there?
[47:56.200 --> 47:57.500]  What was your initial motivation?
[47:57.500 --> 48:04.500]  Well, the frank reason is that I needed a job because I needed money to live and eat food and pay rent.
[48:05.160 --> 48:13.560]  But I mean, I was applying to a number of places, Facebook among them.
[48:13.740 --> 48:16.420]  This was back in 2017 or so.
[48:17.460 --> 48:20.000]  It was actually very silly how I got recruited.
[48:22.340 --> 48:33.300]  I had an acquaintance who I met through a group of ex-astrophysicists in the Bay Area in data science, which is specific enough that it tells you all you need to know about the academic diaspora.
[48:34.580 --> 48:37.020]  And so one of these people worked at Facebook.
[48:37.020 --> 48:38.500]  He randomly recommended me.
[48:38.900 --> 48:39.800]  He never told me anything.
[48:39.960 --> 48:44.220]  I'm assuming he got a bonus for recommending people and just recommended everyone he knew.
[48:45.180 --> 48:46.320]  And so they tried to recruit.
[48:46.580 --> 48:48.200]  They sent me a bunch of recruitment letters.
[48:48.440 --> 48:52.100]  And eventually, I mean, I got semi-downsized from my last job.
[48:52.300 --> 48:55.640]  So I started... and so I responded to one of them.
[48:55.720 --> 48:58.880]  I told them, I told them, hey, let's set up an interview.
[48:58.980 --> 49:00.400]  And they said, okay, let's do it at this time.
[49:00.460 --> 49:01.240]  And I got to that time.
[49:01.360 --> 49:02.180]  I was ready with my phone.
[49:02.260 --> 49:07.300]  I waited and they waited and they called me back a few days later and was like, I'm so sorry.
[49:07.300 --> 49:10.480]  Let me send it to this other person who I swear I want to do the same thing.
[49:11.220 --> 49:12.020]  Except they did.
[49:12.700 --> 49:15.580]  And so at that point, I gave up on them.
[49:15.740 --> 49:22.840]  But a few months later, they messaged me, hello, we see that we have tried to recruit you many times and it has not worked out.
[49:23.120 --> 49:27.160]  It's your anything, just anything we can do to convince your mind otherwise.
[49:29.020 --> 49:31.340]  I told them that they could get better recruiters.
[49:32.760 --> 49:33.800]  I'm glad you're laughing.
[49:35.100 --> 49:37.840]  I'm glad that my deadpan is actually effective.
[49:39.260 --> 49:42.500]  And so anyways, I mean, I'm sure they're looking back at this.
[49:42.660 --> 49:43.660]  This was a dodged bullet.
[49:43.880 --> 49:45.280]  We shouldn't have sent all the last email.
[49:46.200 --> 49:50.320]  But anyways, the interviews went well.
[49:50.520 --> 49:57.680]  I mean, once they actually made the job offer, I was... once they actually made the job offer, I was upfront that I didn't think Facebook was good for society.
[49:58.300 --> 50:00.100]  I told them, that's why I was joining.
[50:00.580 --> 50:02.980]  If there's something that's perfect, there's nothing to improve.
[50:03.180 --> 50:04.360]  Anything you change would make it worse.
[50:06.140 --> 50:09.540]  Which is still sort of my philosophy to a large extent.
[50:11.680 --> 50:16.800]  I was not fond of Facebook when I joined it, and I certainly never expected it to get to this level.
[50:17.300 --> 50:18.960]  If I did, I might have run away screaming.
[50:19.500 --> 50:21.620]  I mean, I'm not a social person.
[50:21.680 --> 50:22.560]  I'm not an extrovert.
[50:22.560 --> 50:27.520]  I hide in... I would prefer to hide in my room with my cats.
[50:27.800 --> 50:28.740]  They're good cats.
[50:31.440 --> 50:36.380]  But... and... but anyway, she told me something like, you'd be surprised how many people at Facebook say that.
[50:37.880 --> 50:49.020]  And like... which I mean... like historically, Facebook was a fairly open company in the sense that people were... people were allowed to express their discontent to the entire company.
[50:49.460 --> 50:51.680]  People could say, I think Facebook should do this better.
[50:52.200 --> 50:58.440]  They had regular surveys about how much... about... to their employees of... do you think Facebook is good for society or not?
[50:59.200 --> 51:04.080]  The number of people responding yes to that was generally in the 50 to 70% range.
[51:04.280 --> 51:05.620]  It was around 70% when I joined.
[51:05.760 --> 51:06.960]  It was 50% when I left.
[51:07.360 --> 51:10.600]  My guess is that it's about 50% now, but I obviously don't know.
[51:11.060 --> 51:14.680]  I mean, there's... there's a bunch of factors that go into it because of self-selection.
[51:15.140 --> 51:18.900]  If you... if you think Facebook is the devil, you're not likely to join Facebook.
[51:19.180 --> 51:23.220]  If you think that Facebook is the greatest thing since sliced bread, you're more likely to join Facebook.
[51:23.480 --> 51:24.960]  The same thing goes for everything else.
[51:25.080 --> 51:28.880]  I mean, black hat hackers tend to be think... people who think black hat is great or something.
[51:30.920 --> 51:49.520]  And so... and so... but... I mean, taking a step back and looking back, I do... I do still stand by my decision in the sense that... I mean, it's nice... it's nice to say everyone should create Facebook, but people have said that many times over the past decade,
[51:49.800 --> 51:51.400]  more than a decade.
[51:52.100 --> 51:59.380]  And in practice, Facebook is part of society going forward, at least in the... at least in the short-term future.
[52:00.260 --> 52:02.560]  I mean, it... I mean, it's hard to look at the long-term.
[52:02.720 --> 52:07.260]  Facebook hasn't died yet, but you could say... but I haven't died yet myself, and I don't think I'm immortal.
[52:08.860 --> 52:12.360]  And so... but... and... and ultimately...
[52:13.620 --> 52:16.240]  like, there are some... there are some positions that I would not join.
[52:16.420 --> 52:32.220]  Like, I can't... I went in from the start, being... being sure that I wanted to join... to go to a place where I was focused on improving the company, where I thought that I would have some sort of net positive impact over generic Facebook employees 354 who would otherwise be in that spot.
[52:33.880 --> 52:36.020]  And... but...
[52:36.680 --> 52:43.320]  I'm not sure I would be comfortable in an actual leadership role at... at a company like Facebook, because there would be so much more politic, you know, etc.
[52:43.620 --> 52:45.560]  But... I stand by my decision.
[52:45.760 --> 52:46.160]  I had...
[52:46.160 --> 52:50.180]  I was able to do a lot more that would not have been done if I hadn't been there.
[52:50.520 --> 52:57.800]  And I... and I... and I... and I would like to... and... I mean, it's to... it's ultimately up to the historians to decide my impact.
[52:57.980 --> 53:00.720]  But I would like to say that I had a decent impact on the world.
[53:01.700 --> 53:09.200]  So, how long did you work there before you started getting the idea that there was something you might eventually want to blow the whistle on?
[53:11.660 --> 53:14.660]  It's hard to say, because there was never any specific moment.
[53:15.020 --> 53:24.080]  Like, if you're reading a book, you're looking at a movie, there's some dramatic moment of tension in which everyone's like, oh, this is the moment of epiphany that changes my life.
[53:24.200 --> 53:25.900]  It doesn't work that way in real life.
[53:27.120 --> 53:38.220]  And so, like, the first time I seriously considered... considered going out and blowing the whistle was... I mean, I think that was probably late... late 2018.
[53:40.260 --> 53:49.180]  I mean, it was not... I mean, I was concerned... I was concerned about... about activity I was finding in the Brazilian... in the Brazilian elections.
[53:49.400 --> 53:51.060]  Brazil had general elections in 2018.
[53:51.060 --> 53:58.860]  2017 that I was... that I eventually was able to get Facebook to actually act and take down, which I... which is the reason that I'm not talking about it right now.
[54:00.260 --> 54:01.780]  And... and... but...
[54:02.880 --> 54:06.900]  I mean, it was in the back of my head... it was in the back of my head for a long time.
[54:08.380 --> 54:11.620]  At... at... at the start, I was hopeful that I was making a difference.
[54:11.920 --> 54:14.020]  And I think I did make a difference in some regards.
[54:14.320 --> 54:20.680]  In terms of concrete difference, I had more impact on Facebook's actions from within the company than everything I've done since speaking out.
[54:20.940 --> 54:27.820]  I mean, I might have had some indirect impacts, but... in the long term, but... I... that's too soon to say at this point.
[54:30.280 --> 54:38.160]  Like... at the start, I was hopeful... because I was... I was hopeful that... okay, I just found this, now I need to yell at people to try and get this changed.
[54:38.340 --> 54:39.700]  Maybe they'll... eventually they'll change it.
[54:39.800 --> 54:42.960]  It's true I'm doing this all in my spare time right now, but they can't keep that up forever.
[54:43.400 --> 54:44.520]  Spoiler alert, they did.
[54:45.580 --> 54:49.920]  And so... and... and... and... like... I thought I was building political capital and stuff.
[54:50.040 --> 54:53.420]  I was talking to a lot of leadership people who I... should not have been talking to.
[54:53.680 --> 55:00.760]  Like I was briefing the vice president on a... Facebook on... of integrity at Facebook on an issue, which like... I was super low level.
[55:00.880 --> 55:02.720]  This is like the equivalent of... I don't know.
[55:03.280 --> 55:04.180]  I... I don't know.
[55:04.380 --> 55:09.940]  Usually you don't get... Usually if you're super low level, you don't get called up to brief the vice president when you've been at the job for a year.
[55:10.540 --> 55:12.200]  But I don't know about other companies.
[55:13.080 --> 55:15.260]  And... and it's not a giant company like Facebook.
[55:15.560 --> 55:21.680]  I mean, my first job... my first job was a tiny startup and I reported to the CEO, somehow marketing to one of the same relationship.
[55:21.680 --> 55:22.440]  I don't know why.
[55:24.380 --> 55:37.160]  So something I didn't realize until I read one of the stories about you is that you... you... met with a reporter from The Guardian 18 months before the actual leaks came out, right?
[55:37.340 --> 55:40.660]  So in that meeting, what did you hope to accomplish?
[55:41.080 --> 55:49.660]  And was there ever the possibility of providing this information without your identity attached to it, essentially blowing the whistle anonymously?
[55:49.660 --> 55:50.720]  Yeah, absolutely.
[55:51.140 --> 55:56.320]  So... so this meeting was... so this meeting was in like... I think November 2019.
[55:56.820 --> 55:57.720]  It was in late 2019.
[55:57.980 --> 55:59.160]  I don't remember the exact date.
[56:00.260 --> 56:03.500]  And it was in... it was in Oakland in California.
[56:03.720 --> 56:04.480]  I live in the Bay Area.
[56:04.640 --> 56:05.640]  I'm just visiting here.
[56:06.840 --> 56:09.280]  Yeah, I was super paranoid about it.
[56:09.460 --> 56:10.400]  I made sure to...
[56:10.400 --> 56:16.120]  I made sure to drop off all my electronics with a friend beforehand who served... who agreed to serve as an alibi.
[56:16.580 --> 56:20.580]  If anyone asked, I was putting cats at their place and taking a nap, which I do plenty.
[56:23.560 --> 56:27.320]  And I put on... I put on a cute dress, which I do not usually wear.
[56:27.460 --> 56:27.660]  I mean...
[56:28.880 --> 56:37.280]  because the idea was... because the idea was that if anyone noticed, hey, this person is sneaking around, they set up an alibi, they're dressed up unusually cute.
[56:37.600 --> 56:40.260]  What is the thought that comes into your head?
[56:40.620 --> 56:41.460]  You're having an affair.
[56:42.100 --> 56:46.360]  And so that was my... that was the alibi to my alibi, which did not end up...
[56:50.800 --> 56:51.160]  which...
[56:51.760 --> 56:52.720]  Oh, thank you.
[56:52.980 --> 56:54.500]  It did not end up being necessary.
[56:54.760 --> 56:57.540]  I mean, no one asked, what are you... what the heck are you doing there?
[56:57.640 --> 56:59.140]  I got there perfectly fine.
[56:59.360 --> 57:10.160]  And the reason I set it up in the first place was that like... was that basically it occurred to me that I was basically... what do you call... is it... do you call it a bus problem or something?
[57:10.520 --> 57:13.480]  If this person gets run over by a bus, everything is toast.
[57:13.840 --> 57:21.920]  And so basically, I made her... I... basically, I told her, don't use any of this unless I die.
[57:24.240 --> 57:25.500]  We joked about it a bit.
[57:25.640 --> 57:28.320]  I think she said, I promise... I promise I won't assassinate you.
[57:28.440 --> 57:30.340]  That would be a silly loophole or something.
[57:33.200 --> 57:34.200]  Or something like that.
[57:34.260 --> 57:34.820]  I don't remember.
[57:35.040 --> 57:35.820]  It's been a while.
[57:38.340 --> 57:46.720]  And... and... and... and... and I... I told her... basically, I was setting up a contingency plan for what happened if I suddenly got run over by a bus.
[57:46.860 --> 57:48.340]  It did not end up being necessary.
[57:49.640 --> 57:59.200]  And... and... and eventually... and... yeah... I... I decided about half a year later that I... that I did want to... that I did want to come forward.
[57:59.200 --> 58:04.520]  But I thought it was my response disability to stay out through... 2020 through the elections and everything.
[58:04.760 --> 58:06.840]  I didn't know how well they were going to get.
[58:08.240 --> 58:11.780]  And... anyways... and... about keeping my identity anonymous.
[58:12.100 --> 58:13.920]  I mean... in principle, that was obvious.
[58:14.040 --> 58:14.760]  Always an option.
[58:14.940 --> 58:18.980]  But in practice, it would never have been anonymous from the company and my co-workers.
[58:19.600 --> 58:21.540]  Because I was the only person who worked on this.
[58:21.620 --> 58:23.040]  And the only person who cared about this.
[58:23.220 --> 58:25.980]  So it would be super obvious to everyone inside the company.
[58:26.240 --> 58:29.600]  If anonymous whistleblower comes... forward with this.
[58:29.760 --> 58:32.360]  And so the question is really, am I going to be anonymous to the public?
[58:32.680 --> 58:36.400]  And... I mean... I mean... that was an option I had.
[58:36.480 --> 58:36.740]  I mean...
[58:37.500 --> 58:38.780]  I had that option too.
[58:39.200 --> 58:44.920]  When BuzzFeed published over my objections, I told them to go ahead with my name because I believe in accountability and everything.
[58:45.800 --> 58:46.900]  But I mean, that's me.
[58:47.280 --> 58:49.420]  There are plenty of people who have made different decisions.
[58:49.600 --> 58:50.640]  I don't begrudge them that.
[58:52.120 --> 59:00.280]  So I'm really fascinated by the mechanics of whistleblowing and the steps people take to protect their anonymity throughout the process.
[59:00.280 --> 59:02.820]  At least up to the point where they actually leak the documents.
[59:03.040 --> 59:08.940]  And I think from what I've read, your process was one of the more careful ones that I've seen people do.
[59:09.220 --> 59:09.840]  It's hard to say.
[59:09.840 --> 59:10.380]  Yeah.
[59:10.520 --> 59:27.460]  Could you talk a little about how you went about finding a trustworthy reporter that you actually wanted to talk to and believed would take the necessary steps to protect you and what you did to protect yourself from your employer finding out that you were doing this?
[59:29.040 --> 59:38.920]  So I was on the very careful side, but it's hard to say how I compared to other people because, again, if you're trying to be anonymous, then if everyone knows about you, then you failed.
[59:38.920 --> 59:44.520]  So I'm sure there are people who have done great who we don't know about because they did great.
[59:46.160 --> 59:53.180]  So anyways, going back, I mean, when I met with Junela, it was half sounding her out for potentially in the future.
[59:53.400 --> 59:59.200]  I mean, I guess I built up a relationship with her over the course of a year and a half or so.
[59:59.840 --> 01:00:02.360]  That made me decide to go forward with her eventually.
[01:00:04.040 --> 01:00:10.740]  Like, there are a lot of different approaches and steps you can take when blowing the whistle.
[01:00:12.840 --> 01:00:14.540]  And whistleblowing is a vague term.
[01:00:14.700 --> 01:00:20.220]  It's used for everything from going with documents to the government to leaking something to the press or et cetera.
[01:00:21.420 --> 01:00:22.680]  And anyways...
[01:00:25.300 --> 01:00:31.760]  I mean, it was obvious to me from the start that I could not come forward completely anonymously.
[01:00:31.960 --> 01:00:34.300]  And that was frankly not of too much interest to me either.
[01:00:34.920 --> 01:00:36.440]  Though I don't know.
[01:00:36.580 --> 01:00:40.660]  I'm not super ensued by the idea of living forever, but maybe I wouldn't say that if I were immortal.
[01:00:45.900 --> 01:00:48.800]  Like, Facebook was going to find out eventually.
[01:00:49.360 --> 01:00:53.860]  The question was, how do I prevent them from finding out until my pace?
[01:00:54.020 --> 01:00:59.780]  And so, I mean, in my case, I mean, different people have different approaches to whistleblowing and leaking and et cetera.
[01:01:00.040 --> 01:01:12.620]  Like, I chose to focus on coming forward with materials that I was a personal expert on, that my work was integral to, that I could be absolutely certain that I had interpreted correctly, that I wasn't taking them out of context, or et cetera.
[01:01:13.420 --> 01:01:16.840]  That I could answer any questions about them.
[01:01:17.360 --> 01:01:18.620]  Different people have different approaches.
[01:01:18.980 --> 01:01:20.740]  Like, I don't begrudge Frances her choice.
[01:01:20.840 --> 01:01:23.740]  I don't begrudge Chelsea her choice, for that matter.
[01:01:24.420 --> 01:01:30.080]  And so, because of that, it was probably less suspicious.
[01:01:30.400 --> 01:01:33.520]  I mean, I was very worried about Facebook security people when I started this.
[01:01:34.000 --> 01:01:41.280]  Like, for instance, when I documented things, first I would make up an excuse to cover up the webcam.
[01:01:41.720 --> 01:01:44.440]  I would go into the bathroom with my laptop, then cover up the webcam.
[01:01:46.640 --> 01:01:49.000]  So, I had an excuse as to why I would do it.
[01:01:49.280 --> 01:01:55.820]  Then I would take pictures, then I would keep it covered, take pictures of the laptop with my personal phone.
[01:01:55.980 --> 01:01:59.340]  So that, I mean, because if you take screenshots and send them the email, obviously.
[01:02:00.020 --> 01:02:01.220]  That's not good.
[01:02:01.220 --> 01:02:06.700]  And they used VPNs and didn't do anything on work networks, and et cetera.
[01:02:08.160 --> 01:02:10.980]  And they only... And I was very careful.
[01:02:11.140 --> 01:02:18.640]  I only opened... Like, even though all these were documents that I worked on and conversations that I was having, I only worked on them when I had an excuse.
[01:02:19.040 --> 01:02:27.180]  And the excuse was that I was looking through them for performance review season in which I needed to tally up what I did and everything.
[01:02:27.400 --> 01:02:29.200]  So that was a convenient excuse to look through everything.
[01:02:29.580 --> 01:02:35.660]  But, I mean, it frankly... I mean, what Frances did was she went, she got a lot of documents.
[01:02:36.180 --> 01:02:40.180]  Frances Haugen, a different whistleblower, who you probably have heard of, who is a lot more famous than me.
[01:02:41.100 --> 01:02:49.480]  And so what Frances did is that she looked up a lot of documents that were public-facing within the company that she didn't work on.
[01:02:49.580 --> 01:02:51.580]  And that was much more sketchy than what I was doing.
[01:02:51.720 --> 01:02:57.600]  And it was, frankly, very surprised that Facebook didn't catch her, because I was terrified the whole time.
[01:02:58.960 --> 01:03:03.080]  But anyways, like, I don't know.
[01:03:03.280 --> 01:03:04.520]  I was very careful.
[01:03:05.620 --> 01:03:14.820]  But it's impossible to say how much is paranoia and how much is justified, which I'm sure is an experience very familiar to people in this room who work on that sort of thing.
[01:03:15.680 --> 01:03:18.520]  You only know that you haven't been paranoid enough when it's too late.
[01:03:20.100 --> 01:03:27.760]  So as far as I can tell, Facebook's official response to all of these things that you've come out with is, we don't know about that.
[01:03:29.420 --> 01:03:31.680]  And it's just sort of this denial.
[01:03:32.080 --> 01:03:36.380]  And are you worried at all about legal action from them against you?
[01:03:36.500 --> 01:03:36.500]  Yeah.
[01:03:37.000 --> 01:03:38.360]  So let's break this down.
[01:03:38.520 --> 01:03:40.000]  First, how do they respond?
[01:03:40.140 --> 01:03:41.260]  Secondly, legal action.
[01:03:42.460 --> 01:03:46.200]  So Facebook has had two general types of response to me.
[01:03:46.200 --> 01:03:51.540]  The first is that it's a very, I don't know what to call it, standard cooperatives.
[01:03:51.860 --> 01:04:01.540]  They gave something that looks like a denial, that is not actually a denial, that is close enough that they look like it is green, but that's not actually engaged on the substance.
[01:04:02.000 --> 01:04:03.500]  I'm going to use an analogy.
[01:04:03.900 --> 01:04:06.080]  So my girlfriend, Lisa, is in the audience.
[01:04:06.400 --> 01:04:14.420]  And suppose that, suppose that tomorrow she asks me, well, when we go back home in California, she asks me, hey, Sophie, did you wash the dishes?
[01:04:14.760 --> 01:04:18.960]  And they look at her and say, I always prioritize washing the dishes.
[01:04:19.280 --> 01:04:23.180]  I have washed the dishes 50 times in the past year.
[01:04:23.360 --> 01:04:28.380]  I have invested over a hundred dollars in cleaning supplies over the course of my life.
[01:04:28.780 --> 01:04:31.100]  Washing the dishes is a deeply difficult problem.
[01:04:31.320 --> 01:04:33.020]  And it will never be fully solved.
[01:04:42.390 --> 01:04:44.850]  And I do want to be clear, I did not make up this analogy.
[01:04:45.170 --> 01:04:48.470]  I bought, I stole it from Issie Lapowsky, who I think is a reporter for Wired.
[01:04:48.610 --> 01:04:50.210]  She gave me permission like a year ago.
[01:04:53.890 --> 01:05:06.550]  And so that's the first type of Facebook denial, in which they look like they're disagreeing, but they're not actually engaging on the substance, because they refuse to engage on the substance, because they don't want to actually answer the actual question.
[01:05:06.830 --> 01:05:11.570]  Like if I said that to my girlfriend, she would be very justified in being upset at me.
[01:05:11.830 --> 01:05:15.130]  But somehow when this company doing it, everyone's like, oh yeah, that's a totally fine answer.
[01:05:16.610 --> 01:05:21.530]  The second response is, they've only given this in one case, which is India.
[01:05:22.370 --> 01:05:25.270]  Very non-star resort, TRDR in India.
[01:05:25.410 --> 01:05:28.930]  In India, I caught a lot of people across the political spectrum.
[01:05:29.130 --> 01:05:43.030]  Facebook agreed to take down all of them, up until we discovered that one of these networks was being run with little obfuscation, out of the account of the sitting Indian parliamentary chairperson of Essex, quite ironically.
[01:05:43.850 --> 01:05:51.190]  And so as soon as that was discovered, I got the stone wall, couldn't get an answer from anyone, even though they had already agreed to take it down.
[01:05:52.290 --> 01:06:02.690]  And so anyways, on that specific case, they've done something very different, which is they changed... they seem to have changed the response so quickly that they hope no one can keep up with it.
[01:06:03.310 --> 01:06:07.610]  First, they gave the Guardian three different responses that you can see in the Guardian's original article.
[01:06:07.770 --> 01:06:09.870]  They said, first, this is a lie, it didn't happen.
[01:06:10.270 --> 01:06:11.670]  Then we showed them the actual documents.
[01:06:11.830 --> 01:06:15.270]  They were like, okay, okay, it did happen, but we took it down.
[01:06:15.710 --> 01:06:16.550]  No, you didn't.
[01:06:17.250 --> 01:06:23.890]  Yes, we took it down with a different team that did not interact with Sophie and did not tell her anything and was not recorded in the documentation.
[01:06:24.610 --> 01:06:26.530]  Which, okay, sure, maybe you did it.
[01:06:26.890 --> 01:06:27.470]  We took it down.
[01:06:28.050 --> 01:06:31.670]  And according to them, they took it down this way six months later.
[01:06:31.810 --> 01:06:33.590]  And I'm like, well, I can't deny this.
[01:06:33.750 --> 01:06:37.110]  They say that they took it down without telling me, without letting myself know.
[01:06:37.230 --> 01:06:38.690]  I mean, why am I to deny that?
[01:06:38.970 --> 01:06:42.850]  And so I was like, okay, I'll be charitable.
[01:06:43.430 --> 01:06:47.070]  When I talk to people, I'll say, according to Facebook, they took it down, but they still waited half a year.
[01:06:47.630 --> 01:06:49.750]  Except that when I told them to them, they were like, no, we didn't.
[01:06:49.790 --> 01:06:50.550]  We didn't take any time.
[01:06:50.710 --> 01:06:52.330]  I'm like, that's exactly what you said.
[01:06:52.870 --> 01:07:16.950]  But, and the headline is like, Sophie, John, and Facebook disagree on what happened, why, et cetera, because... And anyways, I got so fed up with this and going back and forth on testifying to the Indian government that they eventually released the actual documents in question about... I don't remember,
[01:07:17.110 --> 01:07:18.130]  like, 60 pages.
[01:07:18.650 --> 01:07:34.370]  It was... I released them with... I tried to get 30 Indian news outlets on board, 22 agreed, 15 actually published, but it was major news in India for one day before they forgot about it and went to the next new thing.
[01:07:34.930 --> 01:07:44.130]  And anyways, Facebook's official response to my publicly releasing the documents so the entire world could see them was, we cannot comment on this because we have not yet seen the documents.
[01:07:45.550 --> 01:07:47.750]  And so that's where we are right now.
[01:07:47.990 --> 01:07:51.030]  And so that's the second type of response from Facebook.
[01:07:51.410 --> 01:07:54.510]  And they see... It's both sad that...
[01:07:54.510 --> 01:07:58.810]  It's both funny how often they've changed their story and sad that no one's held them to account on it.
[01:07:59.270 --> 01:08:02.150]  And so anyways, that's Facebook's responses.
[01:08:02.410 --> 01:08:06.140]  And in terms of lawsuits, Facebook could sue me for any... for every penny if they wanted to.
[01:08:07.820 --> 01:08:10.760]  I'm protected by that more by the Streisand effect than anything.
[01:08:11.920 --> 01:08:18.020]  Frankly, though, before I came forward, I was much more... I was much more varied about...
[01:08:18.020 --> 01:08:29.600]  about... about... governmental action, both from the fact that I've... that I've made enemies with two world governments that have a track record of their political enemies meeting with... and with suspicious accidents.
[01:08:30.860 --> 01:08:34.500]  And from the fact that what I did could... could allegedly be... be violating under U.S. federal law.
[01:08:37.400 --> 01:08:46.100]  Well, I know you wouldn't want to say this, but I had conversations with my girlfriend before coming forward about what would happen if she were compelled to testify against me, for instance.
[01:08:46.880 --> 01:08:57.080]  Because, I mean, it could be alleged that I violated the trade secret act, which I think it was Anseline Lewandowski, the Uber... the Google Waymo person who was convicted on it.
[01:08:57.200 --> 01:09:02.900]  Because it could be alleged that they took trade secrets from Facebook and... and it could also... and... and...
[01:09:02.900 --> 01:09:08.620]  And it could have also been alleged that they violated the... the CFAA before the Van Buren versus U.S. case.
[01:09:09.280 --> 01:09:12.120]  And so, that was something my lawyer was very concerned about.
[01:09:12.240 --> 01:09:17.180]  I was like, if it comes to that, I'll be... I'll be... I'll do civil disobedience.
[01:09:17.420 --> 01:09:18.400]  I'm fine with that.
[01:09:18.580 --> 01:09:21.700]  But, I mean, thankfully, Facebook is not that popular on Capitol Hill.
[01:09:22.040 --> 01:09:25.120]  And so... and so... and so I'm not... and so I'm...
[01:09:25.120 --> 01:09:26.400]  and so I'm not being arrested.
[01:09:26.700 --> 01:09:28.680]  But, I mean, that was something we discussed in depth.
[01:09:28.860 --> 01:09:33.100]  And it's an example of why, if you want to be a whistleblower, you should talk to a lawyer.
[01:09:34.860 --> 01:09:37.100]  So, I want to leave a lot of time for questions.
[01:09:37.420 --> 01:09:42.920]  So, just a few more here about the consequences that have happened since you did the leaks.
[01:09:43.300 --> 01:09:45.860]  What has the impact been on your personal life?
[01:09:45.980 --> 01:09:46.720]  What do you do now?
[01:09:46.960 --> 01:09:48.800]  And what... what are you planning to do in the future?
[01:09:49.420 --> 01:09:51.780]  Right now, I stay home and pet my cats.
[01:09:51.900 --> 01:09:52.480]  They're good cats.
[01:09:53.000 --> 01:09:56.800]  I also go... I also go to conferences and do interviews and talk to people.
[01:09:57.100 --> 01:09:58.760]  I mean, I... I don't like doing it.
[01:09:58.920 --> 01:10:01.040]  I would much rather stay home and pet my cats.
[01:10:01.180 --> 01:10:03.260]  I'm totally not a crazy cat lady, I swear.
[01:10:05.320 --> 01:10:09.220]  And... but... but... I mean, I don't like doing it, but someone has to do it.
[01:10:09.300 --> 01:10:11.060]  I mean, no one likes waking up at 7 a.m.
[01:10:11.080 --> 01:10:12.780]  to go into the office, but you still do it.
[01:10:12.960 --> 01:10:15.860]  So, I mean, it's same... same... same way in my book.
[01:10:16.660 --> 01:10:20.240]  Like, I didn't intend... I didn't intend to be doing this for so long.
[01:10:20.660 --> 01:10:23.960]  Actually, I thought people would stop caring about me after half a year.
[01:10:24.120 --> 01:10:25.600]  But, I mean, apparently, I'm still doing this.
[01:10:25.620 --> 01:10:32.240]  And it just seems a bit unfair to be joining another company and being like, oh, yeah, I have to leave next... next week for HOPE conference.
[01:10:32.400 --> 01:10:36.580]  And I'm leaving a few months to speak at this conference in India.
[01:10:36.760 --> 01:10:39.520]  And I... I mean, that just seems unfair, you know.
[01:10:40.360 --> 01:10:41.740]  Conflicts of interest and all that.
[01:10:42.360 --> 01:10:44.560]  And so, I stay home and pet my cats.
[01:10:44.980 --> 01:10:47.820]  I... I don't... I don't take payments from people.
[01:10:47.940 --> 01:10:55.480]  I think I asked... I think I told them to either... to either waive the payment... the honorarium for this or donate it to the EFF, which seemed appropriate.
[01:10:56.660 --> 01:10:58.320]  Like, that's usually what I... Oh.
[01:11:03.220 --> 01:11:04.120]  Oh, thank you.
[01:11:05.200 --> 01:11:07.980]  I just... I mean, this is something... I do want to be clear.
[01:11:08.100 --> 01:11:15.320]  This is something I can personally afford to do because, A, I have a lot of savings because I've worked at Facebook, even if... even if I'm told that I was underpaid.
[01:11:15.680 --> 01:11:20.600]  And, B, I have a... I have an amazing girlfriend who is happy for me to be her housewife, although we aren't married.
[01:11:28.340 --> 01:11:32.780]  And the reason I did this is just that, I mean, she's out... she's out for the money.
[01:11:33.060 --> 01:11:33.980]  It's the typical...
[01:11:33.980 --> 01:11:37.340]  it's one of the typical criticisms of whistleblowers.
[01:11:37.460 --> 01:11:43.760]  And I wanted to very strongly avoid that and be able to honestly say I have not taken any penny from anyone to gain from my whistleblowing.
[01:11:43.920 --> 01:11:51.760]  But I do want to be clear, you should not necessarily expect this from other people because that is going to limit whistleblowers to people who are rich and self-sufficient.
[01:11:58.020 --> 01:11:59.740]  Well, I'm just so popular today.
[01:12:00.700 --> 01:12:02.120]  I wasn't expecting that.
[01:12:02.720 --> 01:12:03.160]  So, yeah.
[01:12:03.760 --> 01:12:04.400]  So, yeah.
[01:12:04.540 --> 01:12:07.060]  And that's... that's what I do with my life right now.
[01:12:08.360 --> 01:12:10.400]  Frankly, I miss my cats right now.
[01:12:10.940 --> 01:12:11.940]  They're very good.
[01:12:12.580 --> 01:12:17.640]  I'll be taking the train home after this conference, and I'll pet them a lot after this.
[01:12:19.180 --> 01:12:39.100]  So, I just want to emphasize that what you did was really a tremendous act of bravery and of selflessness because not only are you not taking money to, you know, do any conferences or interviews, you actually turned down a $64,000 severance package from Facebook in order to avoid signing a non-disparagement agreement.
[01:12:40.400 --> 01:12:44.860]  So, I also wanted to ask if you did this again, would you do anything differently?
[01:12:45.260 --> 01:12:49.720]  And do you have any particular advice for future whistleblowers out here?
[01:12:50.580 --> 01:12:58.560]  If I did this again, there were two things I would have... that I think I would have done differently that I think for certain.
[01:12:59.060 --> 01:13:05.280]  I mean, I mean... I mean, if I actually found myself in back in time, first I would freak out, then I would go to... I don't know.
[01:13:05.700 --> 01:13:09.200]  I was trying to figure out if I wanted to go with... to the government with all my future knowledge.
[01:13:09.420 --> 01:13:16.600]  But, I mean, taking the question at its premise, I mean, there are two things I would do differently.
[01:13:16.960 --> 01:13:30.560]  First, I would probably have not published the memo that got leaked, because, I mean, as anyone who's done things in an all-nighter can know, if you write 8,000 words in eight hours in an all-nighter, not all of it will be perfect.
[01:13:30.860 --> 01:13:32.040]  You will make mistakes.
[01:13:32.340 --> 01:13:43.400]  And if someone else, say a reporter, decides to spend their holiday time doing their own all-nighter to write an article about it, the result may not... may have some flaws.
[01:13:43.740 --> 01:13:52.720]  And so... and... and so... there were a number of cases in which this led to a lot of misleading reports and misinformation.
[01:13:53.240 --> 01:13:54.920]  For instance, let's give an example.
[01:13:55.080 --> 01:14:01.040]  I use the word actors to refer to people acting to do something, not in the Hollywood sense.
[01:14:01.560 --> 01:14:05.040]  This quote... this quote was reported without context.
[01:14:05.300 --> 01:14:12.340]  And there is still an article online, if you Google, that's in the Indian press that says, I caught Bollywood red-handed.
[01:14:16.750 --> 01:14:21.010]  Because, I mean, like, I did not know that this was going to... I did not know that this was going to be leaked.
[01:14:21.090 --> 01:14:23.890]  I did not know that this was going to publish to go viral, whatever.
[01:14:24.450 --> 01:14:27.030]  And so anyway, and so... but it did.
[01:14:27.150 --> 01:14:31.530]  And so I had written this, assuming that people who read it wouldn't know what I meant.
[01:14:31.690 --> 01:14:34.370]  And it turned out I was writing for a completely different crowd.
[01:14:35.410 --> 01:14:37.970]  And so... and... yeah...
[01:14:40.050 --> 01:14:45.610]  And so... and so... and so I... I did my best to... I did my best to correct the record.
[01:14:45.650 --> 01:14:49.290]  It was not very good because, I mean, that's the way fact checking works.
[01:14:50.990 --> 01:14:54.450]  Someone posts something gets 10,000 likes, the fact check gets 100.
[01:14:55.710 --> 01:15:04.930]  And so... and so anyways, the other thing I would have done differently is that I would have gotten a PR agent or other support before coming forward.
[01:15:05.170 --> 01:15:06.970]  Because right now I'm doing everything myself.
[01:15:07.310 --> 01:15:10.130]  Like, people send me messages, I book sends, etc.
[01:15:10.230 --> 01:15:11.850]  I do interviews, etc.
[01:15:12.170 --> 01:15:15.850]  And... I mean, it's worked... I mean, it's... it's... it's worked out.
[01:15:16.190 --> 01:15:21.350]  But... but frankly, I have not been very good at it, considering that I'm an introvert who prefers to stay home and pet my cats.
[01:15:21.790 --> 01:15:23.090]  And... and considering the fact...
[01:15:23.090 --> 01:15:31.490]  that... as anyone who's done something can tell you, you're knowing something is not actually an effective strategy for doing something important that you've never done before.
[01:15:33.030 --> 01:15:39.070]  And... I mean, I think I offer... I think I do offer some... I mean, by now... by now this has sort of been my brand.
[01:15:39.650 --> 01:15:41.890]  I don't... I don't have a PR agent.
[01:15:42.230 --> 01:15:45.850]  I don't... I don't... I try to keep things extemporaneous.
[01:15:46.210 --> 01:15:48.030]  I talk to everyone who asks, etc.
[01:15:49.490 --> 01:15:53.530]  Because I think it offers something valuable at this point for me to be in this niche.
[01:15:53.830 --> 01:15:57.370]  Because... I mean, there are some people who look at Frances Haugen and say, she's too prepared.
[01:15:57.610 --> 01:15:59.830]  She has some... she... everything she says is scripted.
[01:16:00.170 --> 01:16:01.770]  You can't... you can't trust her.
[01:16:01.870 --> 01:16:03.450]  And then they look at me and say, she stutters.
[01:16:03.590 --> 01:16:04.390]  She has a bad accent.
[01:16:04.550 --> 01:16:05.770]  She doesn't have anyone supporting her.
[01:16:05.890 --> 01:16:07.010]  Nothing she says is scripted.
[01:16:07.070 --> 01:16:07.890]  You can't trust her.
[01:16:08.150 --> 01:16:12.530]  And there is... and I think it is useful to have that contrast right now.
[01:16:12.710 --> 01:16:15.390]  But if I were doing... and so that's why I keep doing this.
[01:16:15.510 --> 01:16:17.430]  And also because this is essentially my brand.
[01:16:17.650 --> 01:16:18.390]  And I'm fine with it.
[01:16:18.430 --> 01:16:19.230]  And I'm fine keeping it.
[01:16:19.530 --> 01:16:23.370]  But it has also, frankly, held me back in a number of ways.
[01:16:23.490 --> 01:16:27.650]  And if I were doing this again, I would probably have bit the bullet and gotten a PR agent.
[01:16:27.850 --> 01:16:29.830]  I did try, but I was not very successful.
[01:16:33.070 --> 01:16:33.930]  Thank you, Sophie.
[01:16:34.090 --> 01:16:36.030]  That was a fantastic interview.
[01:16:36.190 --> 01:16:38.370]  Let's give her another round of applause, please.
[01:17:03.100 --> 01:17:04.900]  We do have time for some questions.
[01:17:05.220 --> 01:17:09.200]  And people in the audience who would like to ask a question, there's a microphone back there.
[01:17:09.400 --> 01:17:12.180]  And we'll do questions, not discussion or dialogue.
[01:17:12.420 --> 01:17:13.780]  So please try to be concise.
[01:17:14.360 --> 01:17:16.820]  We do have a question that came in from the Matrix chat.
[01:17:17.040 --> 01:17:23.420]  And others are, of course, welcome to send in questions if you're in the room on site or remote to the Matrix chat.
[01:17:23.560 --> 01:17:24.420]  I'm monitoring that here.
[01:17:24.580 --> 01:17:29.700]  So a question that came in a little bit earlier in the talk was, I think, related to fact checking.
[01:17:29.960 --> 01:17:35.060]  It was essentially that Facebook is such a rich company, such a large company.
[01:17:35.560 --> 01:17:41.880]  Do they really have an excuse, or what's the problem with resourcing adequately to address the types of concerns that were identified?
[01:17:42.980 --> 01:17:43.500]  Yeah.
[01:17:43.780 --> 01:17:46.020]  So this goes back to what I said before.
[01:17:46.360 --> 01:17:47.460]  Facebook is a company.
[01:17:47.680 --> 01:17:50.100]  The goal is to make money, not to save the world.
[01:17:50.940 --> 01:17:56.440]  I mean, does Philip Morris, the tobacco company, have the resources to make the cigarettes less addictive?
[01:17:56.780 --> 01:17:57.720]  They probably do.
[01:17:58.100 --> 01:18:05.460]  They probably have the resources to give Medicare a small reimbursement every time someone gets sick with lung cancer and needs to be treated by Medicare.
[01:18:05.960 --> 01:18:08.880]  But they're not going to do that because they're a company.
[01:18:10.000 --> 01:18:11.300]  Their goal is to make money.
[01:18:11.620 --> 01:18:15.760]  And if you think that they're bad for that reason, that's not really a criticism of them.
[01:18:15.760 --> 01:18:18.460]  It's a criticism of capitalism, which a lot of people have said.
[01:18:18.880 --> 01:18:22.420]  But there have not been any great replacements for it thus far.
[01:18:23.220 --> 01:18:36.100]  And so I think it's important to really remember that because, I mean, Facebook, Facebook will say a lot of nice, great things about how we're your friends and we're here to connect the world.
[01:18:36.240 --> 01:18:37.580]  And they might even believe some of it.
[01:18:37.720 --> 01:18:40.960]  But at the end of the day, they're there to make money.
[01:18:41.260 --> 01:18:44.520]  And so do they have the resources?
[01:18:44.940 --> 01:18:45.300]  Certainly.
[01:18:45.580 --> 01:18:46.580]  Are they going to use them?
[01:18:47.120 --> 01:18:47.860]  Why would they?
[01:18:49.980 --> 01:18:50.460]  Thank you.
[01:18:50.520 --> 01:18:51.440]  There's a question in the room.
[01:18:52.100 --> 01:18:59.400]  In your defense, I think a lot of us in this room believe that Ronald Reagan was a bad actor.
[01:19:02.080 --> 01:19:13.520]  But I'm actually confused why you didn't manage to organize a number of, or at least people didn't clump around you who were ethical data scientists within Facebook.
[01:19:13.540 --> 01:19:18.760]  And you ended up the only person in Facebook who was really concerned about this.
[01:19:18.980 --> 01:19:21.400]  And you were expressing it in public within the company.
[01:19:21.640 --> 01:19:26.760]  I would think people would align with you and also give you some cover and plausible deniability.
[01:19:27.540 --> 01:19:31.460]  Or is that just the fate of an introvert who tries to do this kind of thing?
[01:19:32.080 --> 01:19:37.080]  I think part of it is that, as you say, I'm an introvert and terrible at organizing people.
[01:19:37.240 --> 01:19:39.000]  I am not the right person for this.
[01:19:40.400 --> 01:19:45.360]  I mean, I'm sure there are people who would love... I mean, I made my first friend when I was like 17.
[01:19:45.960 --> 01:19:51.300]  Like, when I was younger, like, when clerks would go talk to you in the store, do you need any help?
[01:19:51.380 --> 01:19:52.560]  I would run away from them.
[01:19:52.720 --> 01:19:55.000]  I didn't... I hope I didn't traumatize any of them for life.
[01:19:56.680 --> 01:19:58.840]  So yeah, like, I don't know.
[01:20:01.180 --> 01:20:03.840]  Like, there were people who supported me in principle.
[01:20:04.180 --> 01:20:07.780]  But in terms of the actual work, I was doing it largely by myself.
[01:20:08.000 --> 01:20:09.420]  Because, I mean, people are busy.
[01:20:09.580 --> 01:20:10.420]  There are people at work.
[01:20:10.540 --> 01:20:12.160]  And no one likes to be given extra work.
[01:20:12.380 --> 01:20:12.820]  And work.
[01:20:13.160 --> 01:20:13.580]  And maybe...
[01:20:13.580 --> 01:20:19.960]  And... And... And... And I was never in a position of authority to tell them to do work.
[01:20:20.220 --> 01:20:24.100]  And... I mean, everyone is at different places in their life.
[01:20:24.280 --> 01:20:27.520]  Like, maybe they have a family they have to support, and they can't afford to lose their job.
[01:20:27.600 --> 01:20:29.260]  And so they're not going to spend something on...
[01:20:29.260 --> 01:20:35.980]  that will cause their performance to go down in reviews, because it doesn't do the work that they're required to do.
[01:20:36.820 --> 01:20:42.040]  Like, a lot of... I knew a lot of people who were upset with Facebook, but were there on H-1B visas.
[01:20:42.320 --> 01:20:45.900]  And so their residency was held hostage to their employment.
[01:20:46.600 --> 01:20:50.020]  And so... And so, I mean, it's easy from the outside.
[01:20:50.300 --> 01:20:51.760]  It's easy from the outside to judge.
[01:20:52.040 --> 01:20:58.640]  But, I mean, no one is obligated to torch their life with temporary insanity for an ideal.
[01:21:02.860 --> 01:21:05.900]  Like, joining... Like, the whole Facebook saga was pretty disruptive.
[01:21:06.140 --> 01:21:08.600]  I mean, I lost a lot of friends when I joined it.
[01:21:08.660 --> 01:21:11.780]  I think they thought I was a sellout, but I'm not sure, because I didn't come back afterwards.
[01:21:12.700 --> 01:21:20.760]  I mean, I lost... And of course, a lot of my work... A number of my work people ghosted me after I came forward as a whistleblower, not too surprisingly.
[01:21:21.100 --> 01:21:26.540]  And so... And... But, I mean, and there's also been a pandemic and everything.
[01:21:26.680 --> 01:21:28.180]  I mean, you have lots of data points.
[01:21:28.280 --> 01:21:34.120]  You can't say it's because of this, this, this, unless you know... Unless you've done a careful experiment, which you can't do for this sort of thing.
[01:21:34.340 --> 01:21:42.900]  And so, I guess what I'm really trying to say is, A, it might have been that I could have organized people, but I was frankly very bad at it.
[01:21:43.060 --> 01:21:44.240]  I'm terrible at it.
[01:21:44.400 --> 01:21:45.680]  And so, I failed.
[01:21:45.900 --> 01:21:55.500]  And secondly, there were people who cared, but not enough to, like, actively devote time, that... Because this was not clearly in their areas, the way that it was sort of in mind.
[01:21:55.720 --> 01:22:03.380]  And because it would take a lot of time to look into this, and they would have taken a lot of time for them to essentially do the on-the-job learning that I was.
[01:22:03.480 --> 01:22:04.680]  And I hope that's making sense.
[01:22:06.960 --> 01:22:08.960]  Another question that came in through Matrix.
[01:22:09.180 --> 01:22:11.040]  I think it relates to what you just talked about, actually.
[01:22:11.180 --> 01:22:17.960]  You talked now about some of the emotional toll and some of the collegiality, or lack of it in the workplace.
[01:22:18.180 --> 01:22:21.700]  We had a question that came in about the professional toll that this took.
[01:22:21.900 --> 01:22:25.640]  And if you could talk about that and what some of the costs were professionally.
[01:22:26.280 --> 01:22:33.800]  And also, I think an interesting sideline to that question is if there might have been any professional benefits that you've identified already.
[01:22:34.940 --> 01:22:35.420]  Yeah.
[01:22:36.280 --> 01:22:42.620]  It's hard for me to assess professional toll and benefits because, A, I have very few data points.
[01:22:42.620 --> 01:22:45.000]  This was my second job out of graduate school.
[01:22:45.000 --> 01:22:50.540]  And, B, I haven't been trying to follow up opportunities very currently.
[01:22:52.600 --> 01:22:54.420]  You can tell that I'm a data scientist.
[01:22:55.080 --> 01:22:55.260]  Anyway.
[01:22:56.100 --> 01:22:57.080]  And, like...
[01:22:59.740 --> 01:23:06.800]  So, let's talk about professional impact both within the company while I was doing this, and also after leaving since becoming a vessel blower.
[01:23:07.440 --> 01:23:15.520]  So, within the company, like, Facebook has performance reviews every half year.
[01:23:15.760 --> 01:23:20.160]  And they go from meets no expectations, which is you're fired.
[01:23:20.260 --> 01:23:25.660]  We don't know why we haven't done that already, to redefine expectations, which is we need 10,000 of you.
[01:23:26.120 --> 01:23:29.260]  And so, my reviews were all over the place.
[01:23:29.820 --> 01:23:35.860]  My first half, I was meet some expectations, which is basically shape up where you're fired.
[01:23:36.060 --> 01:23:39.240]  Second half, I got greatly exceeds expectations, which is, oh, my God.
[01:23:40.040 --> 01:23:43.060]  Third half, I... Then I got... Then I got...
[01:23:43.060 --> 01:23:44.780]  Then I got exceeds expectations, which is you're doing great.
[01:23:44.940 --> 01:23:48.080]  Then I got meets most expectations, which is shape up where you're fired.
[01:23:48.180 --> 01:23:48.840]  And I was fired.
[01:23:49.900 --> 01:23:53.360]  And so, this sort of noisiness is not typical.
[01:23:54.820 --> 01:24:03.980]  It's... What's actually most common, I'm told, is that you get meets all expectations since the level for most of your time there.
[01:24:04.940 --> 01:24:12.220]  But I also had three managers during my time at Facebook which I think contributed to that in some regards.
[01:24:12.540 --> 01:24:14.720]  Different ones interpreted my work differently.
[01:24:15.860 --> 01:24:19.940]  And I mean, some of them liked what I was doing, some of them didn't.
[01:24:20.480 --> 01:24:27.740]  I don't think you'll be a spoiler to say that my last manager was not fond of me spending time doing this.
[01:24:29.520 --> 01:24:36.060]  And so it was all over the place with regards to my work at Facebook.
[01:24:36.420 --> 01:24:42.740]  I think they appreciated at first when they were getting a lot of free work because I was essentially doing two jobs at the same time.
[01:24:42.960 --> 01:24:46.160]  But they eventually got annoyed that I was doing it.
[01:24:47.000 --> 01:24:52.400]  With regards to my actual personal professional career outside of Facebook and everything.
[01:24:54.260 --> 01:24:58.920]  Like I have gotten lots of people saying we would like to hire you.
[01:24:59.100 --> 01:25:01.280]  I mean timed with media attention and all that.
[01:25:01.940 --> 01:25:07.720]  It's hard to say how many... I mean, I haven't... I haven't strongly tried to follow up on any of them with one exception.
[01:25:08.640 --> 01:25:14.240]  I haven't tried... I haven't tried to look deeper in them because I'm busy.
[01:25:14.400 --> 01:25:17.840]  And there is some, what do you call it, sampling bias.
[01:25:18.180 --> 01:25:23.860]  If someone thinks that I'm terrible, they are not going to send me great messages about we would like to hire you.
[01:25:23.860 --> 01:25:28.700]  And they are also probably polite enough to not say you are a turncoat who we will never hire.
[01:25:29.080 --> 01:25:34.520]  So I don't know... So I only hear from the people who have strong opinions, which are mostly the people who like me.
[01:25:34.780 --> 01:25:39.280]  And so I don't actually... So it's hard to say because I haven't actually applied for jobs.
[01:25:39.820 --> 01:25:46.200]  But I did get a number... But I did get a number of people who saw my experience as positive and offered to hire me.
[01:25:46.200 --> 01:25:52.460]  I mean, I have heard anecdotally that other prominent whistleblowers have had similar experiences.
[01:25:52.820 --> 01:26:00.700]  And so I would definitely not recommend whistleblowing as a purely selfish move to boost your career and do things.
[01:26:00.880 --> 01:26:08.300]  But I mean, some people have gone that route, which is unfortunately, I think, why we're cynical about whistleblowers these days.
[01:26:10.060 --> 01:26:20.520]  Like, I think last year there was a case in which a Fox News anchor walked off on air saying that she had given a ton of stuff to Project Veritas.
[01:26:20.840 --> 01:26:27.500]  And everyone was very excited up until her evidence was found less than convincing, by which point she had already pocketed a lot of donations.
[01:26:30.360 --> 01:26:40.580]  But, like, in terms of the... I can imagine someone who wants to be famous finding this positive.
[01:26:40.780 --> 01:26:46.500]  But for me, this... But for me, the person also has to be negative when I do it because it's important.
[01:26:47.240 --> 01:26:50.440]  But I mean, it's less negative than working at Facebook was, though.
[01:26:51.720 --> 01:26:52.860]  So, like, I don't know.
[01:26:53.000 --> 01:26:53.320]  Rebound.
[01:26:53.320 --> 01:26:58.100]  Like, because I was frank... Like, I mean, like, I was serious when I wrote that I had blood on my pants.
[01:26:58.260 --> 01:27:04.760]  Like, I started losing sleep at night because I was... Because, I mean, I was staying up and worrying.
[01:27:05.220 --> 01:27:13.080]  Like, I stopped being able to fall asleep reliably without my girlfriend sleeping in the same bed with me while we went dating at the time.
[01:27:13.280 --> 01:27:15.760]  Like many people, I started dating my housemate during the pandemic.
[01:27:17.820 --> 01:27:19.220]  Anyways, I hope that makes sense.
[01:27:19.220 --> 01:27:19.680]  Well, thank you.
[01:27:20.120 --> 01:27:20.620]  Yeah, thank you.
[01:27:20.860 --> 01:27:21.260]  There's...
[01:27:21.260 --> 01:27:27.140]  I know, Jan, if you want to... I don't... It's hard to see people for you, so I'm just helping.
[01:27:27.640 --> 01:27:30.700]  There is another question in the room and we have one queued up in Matrix as well.
[01:27:30.820 --> 01:27:32.300]  So let's hear from the question in the room, please.
[01:27:33.200 --> 01:27:48.260]  So far as I could tell, this sounds like a story of a lot of systematic things where people kind of grasp their hands and say, oh, no, but the structure of the company and all of the environments you are in were really moving against you.
[01:27:48.620 --> 01:28:08.440]  In your time sort of after this or even during, did you have any thoughts on what larger companies could do to try to make sure that when completely unexpected hazards come up like this, they have any structure to be able to deal with them reasonably?
[01:28:09.160 --> 01:28:10.600]  If that question makes sense?
[01:28:11.060 --> 01:28:12.120]  I think so.
[01:28:12.500 --> 01:28:19.540]  So I would... So talk... So I'm going to break this down to what I personally see as a systematic failure that we could...
[01:28:19.540 --> 01:28:22.660]  that we could reasonably expect Facebook to have avoided.
[01:28:24.960 --> 01:28:28.260]  And I'm going to point to two separate categories for this.
[01:28:28.480 --> 01:28:33.500]  The first of them, it's just, for lack of a better word, political interference.
[01:28:34.000 --> 01:28:53.320]  Because at Facebook, the people making decisions about what the rules are on the platform are the same people as those who decide how the rules should be interpreted in prominent cases for the same people who are charged with lobbying politicians and important people and keeping a good relationship with them.
[01:28:53.700 --> 01:29:01.260]  So basically, it's the legislative branch, judiciary branch, and lobbies combined into the policy people who you may have heard of.
[01:29:02.780 --> 01:29:14.180]  And so here in the United States, if a judge is called upon to try a case, and it turns out that they're close friends with the defendant, they might be asked to recuse themselves in favor of a judge who isn't.
[01:29:14.180 --> 01:29:16.140]  At Facebook, it would be the reverse.
[01:29:16.460 --> 01:29:25.160]  A judge who didn't know the defendant would be called upon to give it over to someone who knew the defendant, who was friends with him, ideally who had to play golf with them every week so that they could know them better.
[01:29:25.560 --> 01:29:29.120]  And this is very atypical, as I understand it.
[01:29:30.640 --> 01:29:32.620]  That Facebook has this organization.
[01:29:33.100 --> 01:29:35.000]  I'm told that Twitter doesn't do this.
[01:29:35.320 --> 01:29:38.020]  I haven't worked at Twitter, so I don't know personally this is what I'm told.
[01:29:38.140 --> 01:29:43.000]  That Twitter keeps their lobbying department and they decide whether the rules department are separate.
[01:29:43.000 --> 01:29:47.660]  Facebook is a company, but so is, for instance, the New York Times, the Washington Post.
[01:29:48.220 --> 01:29:53.380]  And these generally do not make a common practice of killing a story because the politician complained.
[01:29:53.880 --> 01:29:58.900]  At Facebook, that happens in analogy so often that it's day to day.
[01:29:59.360 --> 01:30:02.800]  And so, I mean, I'm not the only person to have pointed this out.
[01:30:02.800 --> 01:30:06.940]  I think Alex Damos, Sumit Chakrabarty, have done so as well.
[01:30:07.160 --> 01:30:10.500]  And so, this is a core problem in how Facebook was organized.
[01:30:12.440 --> 01:30:18.760]  Like in some cases, I think it was probably one of the key issues in my work in India.
[01:30:18.760 --> 01:30:26.520]  In other cases, I got pushed back because I was accused of being...
[01:30:28.400 --> 01:30:37.100]  Because people were like, we should protect the rights of the users by not cracking down on them right away and giving them warnings and et cetera.
[01:30:37.640 --> 01:30:46.700]  And that is based on a dynamic in which the police officer is expected to be friends with the person that they're charged with overseeing.
[01:30:47.580 --> 01:30:54.420]  The second point that I would point to is that, for lack of better work, Facebook is too quantitatively focused.
[01:30:54.920 --> 01:30:57.800]  They focused, like many tech companies, they're very number driven.
[01:30:58.100 --> 01:30:59.700]  They like their metrics, their APIs...
[01:31:00.960 --> 01:31:03.220]  Sorry, their KPIs, whatever.
[01:31:04.940 --> 01:31:08.520]  You can tell it's been involved since I have actually worked as a data scientist.
[01:31:13.460 --> 01:31:16.680]  And people like being able to report that...
[01:31:16.680 --> 01:31:23.280]  In my work in this company, I have increased our core metric by 10.6% or something like that.
[01:31:23.540 --> 01:31:28.580]  The flip side of that is that when you focus on that, you do not focus on what cannot be measured.
[01:31:28.580 --> 01:31:37.360]  And that is a lot of the issue that I think Facebook is having, that it is focused on reactivity rather than proactivity.
[01:31:37.580 --> 01:31:46.380]  Because in this analogy, they're grading a fire department on how many fires they put out, without considering how many fires were prevented in the first place.
[01:31:46.820 --> 01:31:49.860]  Like, fire departments used to be paid on how many fires they put out.
[01:31:49.860 --> 01:31:52.480]  I think some of them started setting fires at that point.
[01:31:53.920 --> 01:32:05.880]  And, like I was told, I mentioned that I was told that this wasn't important to Facebook, because if it works, it would blow up and become the news, and that would cause Facebook to pay attention.
[01:32:06.160 --> 01:32:12.080]  And that reflects a fundamentally short-sighted bias that doesn't try to focus on prevention.
[01:32:12.080 --> 01:32:15.700]  Like, Facebook has really stopped trying to prevent bad news cycles.
[01:32:15.860 --> 01:32:16.820]  I mean, it's gotten used to them.
[01:32:17.000 --> 01:32:22.340]  Like, there's so many bad news cycles that they don't even try to stop them anymore.
[01:32:22.440 --> 01:32:29.980]  But frankly, I was surprised when she came forward that Frances wasn't caught looking through documentation.
[01:32:30.720 --> 01:32:37.500]  And I think that's because Facebook has a lot of very smart security people who look and catch people who leak to the press.
[01:32:37.940 --> 01:32:43.580]  The reason I personally think she wasn't caught is that those people, like the rest of the company, are not proactive.
[01:32:43.780 --> 01:32:45.420]  They are focused on their quantitative measures.
[01:32:45.720 --> 01:32:51.100]  And they aren't focused on preventing leaks before they happen because you can't prove them and you don't get rewarded for that.
[01:32:51.560 --> 01:32:56.640]  And so, they're great at finding leaks after they have happened, but she just got everything and then came forward.
[01:32:56.780 --> 01:32:58.280]  And then they're like, well, what can we do?
[01:32:58.500 --> 01:32:59.600]  I mean, at least that's my guess.
[01:32:59.700 --> 01:33:00.480]  I can't read minds.
[01:33:00.860 --> 01:33:05.040]  And so, those are the two points I would point to internally within the company.
[01:33:05.040 --> 01:33:10.920]  With the broader structural issue, the one that they pointed to was the information gap.
[01:33:11.680 --> 01:33:13.620]  That this is really...
[01:33:13.620 --> 01:33:18.840]  An economist would call it an externality problem coupled with information asymmetry.
[01:33:19.320 --> 01:33:24.600]  In layman's terms, the costs of this are paid by society, but not by Facebook.
[01:33:25.140 --> 01:33:26.720]  And society doesn't know about this.
[01:33:26.860 --> 01:33:27.620]  Only Facebook does.
[01:33:27.620 --> 01:33:42.240]  And this is why my testimony... when I testified, for instance, to the British Parliament and etc., I suggested that they focus on red team style penetration tests so that they could try and actually hold the company accountable by knowing what's actually going on.
[01:33:42.500 --> 01:33:43.380]  Hope that makes sense.
[01:33:44.560 --> 01:33:45.560]  Yes, thank you.
[01:33:47.000 --> 01:33:53.080]  Another question from Matrix would like you to dwell further into the emotional toll here.
[01:33:53.320 --> 01:34:00.020]  And it's asking whether you have any resources or techniques that you use to deal with facing some of these, you know, impacts.
[01:34:00.940 --> 01:34:07.500]  I use the techniques and resources of petting my cats, having my cat on my lap, and cuddling my girlfriend.
[01:34:07.500 --> 01:34:08.300]  She's great.
[01:34:08.560 --> 01:34:09.540]  I mean, I don't know.
[01:34:11.180 --> 01:34:12.120]  Excellent answer.
[01:34:12.120 --> 01:34:16.500]  I mean, I used to talk to a therapist.
[01:34:16.820 --> 01:34:35.520]  I mean, I stopped talking to them halfway through this started going on just because either I couldn't tell her about what was happening, or because I would feel like I was, I don't know, I was afraid that she would think that I was delusional or something because it sounds a bit that way.
[01:34:36.180 --> 01:34:38.840]  Like Margaret Mitchell effect or something like that.
[01:34:40.160 --> 01:34:51.400]  Margaret Mitchell was the wife of a cabinet member during the Nixon administration who was put in a psychiatric ward after she said the president's people were following her and eavesdropping on her.
[01:34:52.240 --> 01:34:56.240]  And I might have forgotten her name.
[01:34:56.460 --> 01:35:00.680]  Anyways, I thought about getting a therapist again.
[01:35:00.680 --> 01:35:07.040]  But I mean, I don't do well with, I don't like video calls and et cetera.
[01:35:07.440 --> 01:35:10.700]  And I'm not super comfortable with it in person right now.
[01:35:10.820 --> 01:35:13.840]  So I keep procrastinating on it, I suppose.
[01:35:14.600 --> 01:35:19.740]  And yeah, like, I have talked to some visible aid organizations.
[01:35:20.500 --> 01:35:21.900]  Some have been more helpful than others.
[01:35:24.160 --> 01:35:31.660]  But everything I do, like, everything I do is basically, like, in terms of actual organization, responding to people, doing interviews, et cetera.
[01:35:31.880 --> 01:35:33.900]  That's just by myself, without support.
[01:35:34.740 --> 01:35:36.660]  And as they count to my girlfriend and my cats, they're great.
[01:35:38.420 --> 01:35:40.040]  There's a question in the room, also.
[01:35:41.180 --> 01:35:43.900]  I have a technical question about your methodologies.
[01:35:43.900 --> 01:35:51.840]  You relied on metadata, internal metadata, about, I presume, registrations and behavior.
[01:35:52.200 --> 01:35:57.860]  But did you use any stylometrics on the content of the postings?
[01:35:58.060 --> 01:36:04.360]  There was the leverage effect caused by additional likes and, you know, all of that leverage stuff.
[01:36:04.360 --> 01:36:15.520]  But if there was actual independent content, did you try to analyze it for similarity clustering or, you know, central sources attributing it?
[01:36:17.000 --> 01:36:17.320]  Okay.
[01:36:17.760 --> 01:36:19.360]  So I'm going to try to break that down.
[01:36:19.560 --> 01:36:25.020]  Firstly, so I relied on metadata that did not look into content.
[01:36:25.260 --> 01:36:28.740]  I think you asked about doing similarity analysis on content.
[01:36:29.280 --> 01:36:33.760]  That is something that I did not personally do, but people within the company have done.
[01:36:34.780 --> 01:36:39.240]  This is, like most machine learning or AI, not necessarily super effective.
[01:36:39.820 --> 01:36:42.860]  Because, I mean, sure, you can think it's obvious.
[01:36:43.080 --> 01:36:43.920]  If everyone starts...
[01:36:43.920 --> 01:36:48.000]  If a lot of people start posting this string all at the same time, it might be suspicious.
[01:36:48.760 --> 01:36:51.880]  Here's an example of a string that was a false positive.
[01:36:52.300 --> 01:36:53.620]  People saying, Happy Thanksgiving.
[01:36:55.980 --> 01:37:00.880]  And so, like, this was an actual example from, like, five years ago or something.
[01:37:01.100 --> 01:37:02.320]  It was before my time.
[01:37:02.620 --> 01:37:05.240]  And so this is something that people within the company do.
[01:37:06.160 --> 01:37:08.660]  I do not work on a team or work very closely.
[01:37:08.920 --> 01:37:15.540]  But sometimes, like, for instance, when Facebook takes down links that are false positives, it's this in action.
[01:37:16.720 --> 01:37:31.140]  Because with hundreds of countries and multiple hundreds of languages, you'd need native language expertise in many of these languages to actually do content analysis.
[01:37:31.620 --> 01:37:36.620]  So it's a real problem for every content provider, every platform.
[01:37:37.460 --> 01:37:37.900]  Yes.
[01:37:38.360 --> 01:37:45.900]  So I do want to speak to, with regard to fake accounts and authenticity questions, you don't actually need to know the content very much.
[01:37:46.800 --> 01:37:49.140]  It's a question of, is this person real?
[01:37:49.380 --> 01:37:49.680]  Are they...
[01:37:51.040 --> 01:37:54.660]  And that's something you can figure out perfectly fine with metadata.
[01:37:55.280 --> 01:37:58.960]  And looking at the profile, most of the time, that is content agnostic.
[01:37:59.460 --> 01:38:01.860]  I mean, you can look at the content to see what are they doing.
[01:38:02.300 --> 01:38:05.140]  Like, that can be helpful for context, but it is not important.
[01:38:05.660 --> 01:38:07.280]  In the same way that...
[01:38:08.020 --> 01:38:09.320]  In the same way that...
[01:38:09.320 --> 01:38:10.480]  In the same way that, I don't know, if you're looking for...
[01:38:10.480 --> 01:38:15.460]  If you're looking at ballot box stuffing, you don't need to know who the vote was cast for to know that it's a fake ballot.
[01:38:18.020 --> 01:38:19.400]  And so this...
[01:38:19.400 --> 01:38:20.880]  I mean, I certainly don't speak a Zeri.
[01:38:21.120 --> 01:38:23.780]  I can speak one phrase in Spanish, which is...
[01:38:24.820 --> 01:38:26.560]  Which means, I am petting my cat.
[01:38:33.860 --> 01:38:36.960]  And I relied upon Google Translate and Wikipedia.
[01:38:37.400 --> 01:38:39.600]  At times, I actually considered this an advantage.
[01:38:39.880 --> 01:38:40.800]  Because I was...
[01:38:40.800 --> 01:38:44.060]  Because I was certain that I was not being politically biased.
[01:38:44.280 --> 01:38:46.940]  Because I did not need to know who this person was.
[01:38:47.120 --> 01:38:50.740]  I had no clue who they were, and I was making my decisions regardless of that.
[01:38:50.740 --> 01:38:51.600]  It was...
[01:38:51.600 --> 01:38:54.580]  And all I needed to know was, are these accounts fake?
[01:38:54.780 --> 01:38:56.360]  Are the people supporting politicians?
[01:38:56.640 --> 01:38:56.960]  Yes?
[01:38:57.260 --> 01:38:57.560]  Good.
[01:38:57.760 --> 01:38:58.120]  Done.
[01:38:58.780 --> 01:38:59.840]  And I hope that makes sense.
[01:38:59.840 --> 01:39:00.280]  Yeah.
[01:39:00.720 --> 01:39:03.080]  But my related question has to do with the page.
[01:39:03.240 --> 01:39:04.280]  Concept of page.
[01:39:05.220 --> 01:39:05.600]  In Facebook.
[01:39:06.120 --> 01:39:06.880]  Creative page.
[01:39:09.480 --> 01:39:12.680]  Do you have an independent identity as that page?
[01:39:13.100 --> 01:39:20.080]  So the way that pages work on Facebook is that a user controls that page.
[01:39:20.260 --> 01:39:22.860]  So for instance, I don't have a Facebook account anymore.
[01:39:22.980 --> 01:39:23.600]  But suppose I do.
[01:39:23.860 --> 01:39:24.740]  I'm Sophie Zhang.
[01:39:25.040 --> 01:39:25.740]  I control a page.
[01:39:26.200 --> 01:39:29.240]  Yang very graciously gives me control of the public page.
[01:39:29.240 --> 01:39:33.900]  I am not Yang, but they have given me permission to post things with Yang's public page.
[01:39:34.020 --> 01:39:36.620]  And they also do things with, I don't know, someone else who does it.
[01:39:36.820 --> 01:39:37.520]  And that's fine.
[01:39:37.700 --> 01:39:39.760]  A single person can control a lot of pages.
[01:39:40.500 --> 01:39:41.680]  I mean, often times.
[01:39:41.820 --> 01:39:48.040]  Like maybe there's a page for HOPE, and there's a page for New HOPE 2022.
[01:39:48.540 --> 01:39:52.400]  I mean, it would be reasonable for the same person to control both of these pages.
[01:39:53.340 --> 01:39:55.200]  And that's the intended purpose.
[01:39:55.200 --> 01:40:06.200]  The side effect is that this is also a great, this is also a super easy way to not, to switch between SOS and fake assets without even needing to bother to dock in and out.
[01:40:08.660 --> 01:40:12.480]  There's a, I think a last question here, and then we'll wrap up for the session.
[01:40:12.600 --> 01:40:13.960]  You might have a few closing remarks.
[01:40:14.180 --> 01:40:15.480]  This also came in through Matrix.
[01:40:15.480 --> 01:40:17.120]  It's a little related to what we were just talking about.
[01:40:17.560 --> 01:40:21.460]  And it paints a dichotomy that, first of all, I'd like to see if you agree with this dichotomy.
[01:40:21.560 --> 01:40:26.280]  But it says there's a tension between privacy and authenticity.
[01:40:26.440 --> 01:40:31.680]  The idea being that you need to ask more and more potentially private information, which I think you just refuted.
[01:40:31.820 --> 01:40:35.620]  But you have to ask more and more private information to figure out who someone is.
[01:40:36.100 --> 01:40:37.860]  And there's this tension.
[01:40:37.860 --> 01:40:40.400]  So do you agree with that dichotomy?
[01:40:40.520 --> 01:40:42.060]  And how do you see resolution?
[01:40:42.440 --> 01:40:43.900]  Or do you see that there's a...
[01:40:43.900 --> 01:40:47.220]  I think the real question from Matrix is what maybe is the right balance?
[01:40:47.560 --> 01:40:48.460]  Yeah, absolutely.
[01:40:48.920 --> 01:40:50.680]  So I think it's a partial level.
[01:40:51.260 --> 01:40:54.900]  This dichotomy, because it's partially, it's my personal belief.
[01:40:55.740 --> 01:40:58.960]  I mean, like anonymity by itself doesn't...
[01:40:58.960 --> 01:41:06.560]  It isn't an issue for the classical style of fake accounts.
[01:41:06.700 --> 01:41:11.640]  It's fine if you have one account, if you have multiple accounts, if they don't interact with each other.
[01:41:11.840 --> 01:41:20.660]  If you have one account that does personal things, and one account that does professional things, that is in theory, theoretically, philosophically fine.
[01:41:21.200 --> 01:41:27.000]  But where you start getting into issues is if you start having the accounts interact with each other and pretend to be separate people.
[01:41:27.000 --> 01:41:43.360]  So you have your accounts engaging in a conversation on someone else's, where you pretend to convince someone who's actually just yourself, if you have them start sharing each other's content, if you start having them... if you start having them, I don't know,
[01:41:44.920 --> 01:41:53.420]  agree with the same thing and post the same thing, I mean, that does also create your anonymity because you're doing the same thing with them.
[01:41:53.420 --> 01:41:57.760]  But it would also make this increasingly bad.
[01:41:58.360 --> 01:42:06.140]  The area that I would say is that they do bad up against each other somewhat has to do with enforcement.
[01:42:07.000 --> 01:42:21.380]  So usually at Facebook, the standard response, if you're almost certain that an account is fake, is that you send it to a bunch of checklists in which you require them to provide identification.
[01:42:21.980 --> 01:42:28.900]  So when people say Facebook wanted my passport, Facebook wanted my ID, this is what happens because, I mean, ML is not perfect.
[01:42:29.080 --> 01:42:29.800]  It makes mistakes.
[01:42:30.080 --> 01:42:33.600]  If you don't have any false positives, that means you're not being aggressive enough.
[01:42:36.620 --> 01:42:45.000]  And so this does have the advantage of, because Facebook has a real name policy, it can afford to do this.
[01:42:45.380 --> 01:42:48.300]  In a way that might not quite work for Twitter.
[01:42:48.440 --> 01:42:50.720]  I don't know how Twitter does this, so I'm speculating here.
[01:42:50.720 --> 01:43:04.600]  But the point is that this allows Facebook to do things with greater precision than would be possible if they did not have the ability to ask people for profile, for IDs, etc.
[01:43:04.600 --> 01:43:13.300]  Just because otherwise, there would be more people, either more people incorrectly swept up, or they would take down fewer fake profiles.
[01:43:14.240 --> 01:43:20.740]  And so that's an example of where the dichotomy and tension lies.
[01:43:21.800 --> 01:43:27.320]  Like, there have been a lot of discussions about privacy.
[01:43:27.740 --> 01:43:32.020]  Like, I believe Elon Musk has advocated for authenticating everyone on Twitter.
[01:43:33.300 --> 01:43:43.380]  I mean, a lot of this is tied to the stereotype that fake accounts are responsible for a large proportion of hate speech, misinformation, etc.
[01:43:43.380 --> 01:43:46.320]  Which, like most stereotypes, is completely incorrect.
[01:43:47.400 --> 01:43:56.860]  Like, I mean, unfortunately, most people who share hate speech and misinformation are people who genuinely believe it.
[01:43:57.300 --> 01:44:00.200]  Like, countries have experimented with authenticating everyone.
[01:44:00.780 --> 01:44:03.600]  South Korea did it for a number of years.
[01:44:03.880 --> 01:44:08.140]  It reduced the number of hateful content by, like, maybe 5%, I think.
[01:44:08.140 --> 01:44:15.780]  And it also resulted in the leak of most South Koreans' personal information when the companies that were storing their personal information got hacked.
[01:44:18.040 --> 01:44:24.860]  And so, requiring people to authenticate themselves, etc., would solve exactly one problem, which is fake accounts.
[01:44:25.660 --> 01:44:28.940]  Which, I mean, I think a lot of people have one-size-fits-all solutions.
[01:44:29.720 --> 01:44:36.660]  There's also the notion that breaking up Facebook would solve everything when it would solve precisely one problem, which is that Facebook is too powerful.
[01:44:37.960 --> 01:44:40.620]  And I seem to have gone off on a bit of a segue here.
[01:44:40.760 --> 01:44:42.020]  But I hope this made sense.
[01:44:42.880 --> 01:44:43.440]  Thank you.
[01:44:43.540 --> 01:44:44.660]  We're at the end of our time.
[01:44:44.860 --> 01:44:46.520]  I'd like to mention two things.
[01:44:46.660 --> 01:44:48.120]  One is that our guests are here.
[01:44:48.220 --> 01:44:50.000]  They're participating in the event.
[01:44:50.260 --> 01:44:54.660]  They're happy to speak with you politely and, you know, as time allows.
[01:44:54.900 --> 01:45:00.160]  And also, I think we'll be available just outside the hall if there's a little further discussion anyone would like to have.
[01:45:00.160 --> 01:45:07.360]  I'd like to turn it over to both of you if you have any closing remarks or comments or things that you didn't get to say in the last bit of time together.
[01:45:08.420 --> 01:45:09.200]  Hack the planet.
[01:45:09.980 --> 01:45:10.120]  Go on.
[01:45:10.980 --> 01:45:11.200]  Yeah.
[01:45:11.380 --> 01:45:11.480]  All right.
[01:45:25.050 --> 01:45:25.570]  All right.
[01:45:25.750 --> 01:45:26.570]  Thank you so much.
[01:45:26.690 --> 01:45:28.350]  It's been a real pleasure having you on the HOPE stage.
[01:45:33.920 --> 01:45:34.620]  All right.
[01:45:49.980 --> 01:45:51.500]  Gilman who was on Poseidon?
[01:45:52.020 --> 01:45:53.180]  Oh okay.
[01:45:54.000 --> 01:45:57.900]  Yeah and also you weren't watching to meet you so that was a great...
[01:45:57.900 --> 01:45:59.000]  Thank you.
[01:45:59.000 --> 01:45:59.180]  Thank you.