1 00:00:26,640 --> 00:00:30,360 We know that what we're doing matters to a lot of people. 2 00:00:30,360 --> 00:00:32,600 It matters very much to us. 3 00:00:32,600 --> 00:00:35,240 And, at our scale and our volume, 4 00:00:35,240 --> 00:00:38,720 even if we get it right 99% of the time, 5 00:00:38,720 --> 00:00:41,560 that still means we're making thousands of mistakes every day. 6 00:00:41,560 --> 00:00:45,800 And every one of those mistakes is important to somebody. 7 00:00:45,800 --> 00:00:47,840 For a company called Facebook, 8 00:00:47,840 --> 00:00:50,880 you rarely get to see the faces behind the app. 9 00:00:52,040 --> 00:00:53,560 These are the people in charge 10 00:00:53,560 --> 00:00:56,400 of the biggest social network in the world. 11 00:00:56,400 --> 00:01:00,480 And, with nearly a third of all humans plugged into their platform, 12 00:01:00,480 --> 00:01:03,280 it's an experiment that comes with an enormous risk. 13 00:01:05,440 --> 00:01:08,240 This is the world's brains being networked together 14 00:01:08,240 --> 00:01:10,000 for the first time. 15 00:01:10,000 --> 00:01:11,880 And so there's going to be new issues. 16 00:01:11,880 --> 00:01:15,280 There's going to be new problems and we're also going to make mistakes. 17 00:01:17,640 --> 00:01:18,720 For the first time, 18 00:01:18,720 --> 00:01:22,800 Facebook are letting cameras inside their global organisation. 19 00:01:24,200 --> 00:01:26,640 All right, everybody, we've got an amber alert activation. 20 00:01:26,640 --> 00:01:28,880 Steve, you on me? Yeah. Time 15.20. 21 00:01:28,880 --> 00:01:30,640 All right. 22 00:01:30,640 --> 00:01:32,120 A technological playground 23 00:01:32,120 --> 00:01:34,960 where some of the smartest engineers in the world 24 00:01:34,960 --> 00:01:38,560 are being hired to build something no-one has built before. 25 00:01:38,560 --> 00:01:40,440 When we give feedback, we give feedback in a way 26 00:01:40,440 --> 00:01:41,560 that is constructive. 27 00:01:41,560 --> 00:01:43,480 But we don't believe in being arseholes. 28 00:01:43,480 --> 00:01:45,720 And that's like... It's straight up... 29 00:01:45,720 --> 00:01:47,480 We just... We don't tolerate dicks. 30 00:01:47,480 --> 00:01:48,680 But in the last year, 31 00:01:48,680 --> 00:01:51,640 their experiment to connect everyone on the planet 32 00:01:51,640 --> 00:01:53,920 has gone dramatically wrong. 33 00:01:53,920 --> 00:01:55,680 I think it's time to ask whether 34 00:01:55,680 --> 00:01:59,280 Facebook may have moved too fast and broken too many things. 35 00:02:01,000 --> 00:02:04,440 Data leaks, fake news and hacks on user security 36 00:02:04,440 --> 00:02:08,200 are threatening to destroy everything they've built... 37 00:02:08,200 --> 00:02:12,240 Facebook's interference in our elections and in our society 38 00:02:12,240 --> 00:02:14,600 has de-stabled our democracies. 39 00:02:16,240 --> 00:02:19,240 ..as Facebook has suffered a year of deepening scandals 40 00:02:19,240 --> 00:02:20,880 and intense media scrutiny. 41 00:02:22,280 --> 00:02:25,120 This is the story of how it works, 42 00:02:25,120 --> 00:02:27,200 what they're doing to fix it, 43 00:02:27,200 --> 00:02:31,240 and how Facebook are quietly planning to grow even bigger. 44 00:03:02,080 --> 00:03:04,760 This is the nerve centre of Facebook... 45 00:03:05,800 --> 00:03:09,440 ..where all data on the platform is processed and stored. 46 00:03:10,440 --> 00:03:12,960 Very few outsiders have ever been in here. 47 00:03:14,280 --> 00:03:17,160 So you're talking full distance so we can't see it right now. 48 00:03:17,160 --> 00:03:20,200 But as a store opens you may be able to see all the way down. 49 00:03:20,200 --> 00:03:22,760 But from one end of the building to the other, 50 00:03:22,760 --> 00:03:26,080 you're talking about four American football fields. 51 00:03:31,080 --> 00:03:35,440 This is just 1 of 15 data centres around the globe. 52 00:03:35,440 --> 00:03:38,880 Connected together, they make the Facebook cloud. 53 00:03:40,240 --> 00:03:44,320 At the end of this corridor, 2.3 billion people can share 54 00:03:44,320 --> 00:03:45,800 anything with anyone. 55 00:03:47,600 --> 00:03:50,240 It's one of the biggest anthropological experiments 56 00:03:50,240 --> 00:03:51,680 of the 21st century. 57 00:03:54,520 --> 00:03:56,320 We're about to step into the cloud. 58 00:03:56,320 --> 00:03:59,760 It's hundreds of thousands of servers that store your likes 59 00:03:59,760 --> 00:04:02,840 and your memories that you choose to share with friends and family. 60 00:04:02,840 --> 00:04:04,480 It's not just a data centre. 61 00:04:04,480 --> 00:04:07,200 It's not just data that's stored here. 62 00:04:07,200 --> 00:04:08,760 These are people's memories, 63 00:04:08,760 --> 00:04:11,160 and, you know, we take that very seriously. 64 00:04:24,560 --> 00:04:28,120 Everything you do on Facebook is recorded here. 65 00:04:28,120 --> 00:04:31,680 Worldwide, some 14 billion photos alone are uploaded 66 00:04:31,680 --> 00:04:33,040 every single week. 67 00:04:35,800 --> 00:04:39,120 These machines decide what appears in your news feed 68 00:04:39,120 --> 00:04:41,440 by analysing what you've looked at before. 69 00:04:43,920 --> 00:04:46,720 You go to the Facebook website and you see your news feed. 70 00:04:46,720 --> 00:04:48,760 You think, you know, "How complicated can that be?" 71 00:04:48,760 --> 00:04:51,120 But whatever you choose to click on 72 00:04:51,120 --> 00:04:52,440 or to look at for a while 73 00:04:52,440 --> 00:04:55,880 or to respond to is used to train a machine learning system, 74 00:04:55,880 --> 00:04:57,880 which is basically a model of your taste. 75 00:05:01,440 --> 00:05:04,920 And so that forms your alter ego 76 00:05:04,920 --> 00:05:08,440 in the Facebook servers to figure out what to show you next. 77 00:05:10,960 --> 00:05:14,560 These learning algorithms use this model of your taste 78 00:05:14,560 --> 00:05:19,000 to match you to content you're more likely to be interested in. 79 00:05:19,000 --> 00:05:22,880 Your view is not the same as any of the other billions of people. 80 00:05:22,880 --> 00:05:24,480 It is custom built. 81 00:05:26,560 --> 00:05:29,800 Let's go find all the different latest updates from your friends. 82 00:05:29,800 --> 00:05:31,840 Let's go to get all the different videos. 83 00:05:31,840 --> 00:05:35,600 Let's look at things you might be interested in. 84 00:05:35,600 --> 00:05:40,160 This is happening billions, if not trillions, of times a day. 85 00:05:42,160 --> 00:05:45,080 This matching process is how they make their money. 86 00:05:45,080 --> 00:05:48,480 Selling to advertisers the ability to reach you. 87 00:05:49,960 --> 00:05:54,360 Your data is anonymised but you are the product. 88 00:05:54,360 --> 00:05:58,640 It's how they made $55.8 billion in revenue last year alone. 89 00:06:00,840 --> 00:06:03,080 This is one of the largest vaults of personal data 90 00:06:03,080 --> 00:06:04,920 anywhere on the planet. 91 00:06:04,920 --> 00:06:07,680 And that comes with a moral responsibility. 92 00:06:10,200 --> 00:06:13,640 But this is a long way from what its creators first imagined. 93 00:06:22,200 --> 00:06:25,240 When Facebook launched in 2004, 94 00:06:25,240 --> 00:06:28,280 it was just a website for college kids to stay in touch. 95 00:06:29,920 --> 00:06:31,400 For more than a decade, 96 00:06:31,400 --> 00:06:34,200 Mark Zuckerberg and a team of engineers 97 00:06:34,200 --> 00:06:37,240 tried to get as many people to join as possible. 98 00:06:38,680 --> 00:06:41,080 They took risks, embraced new approaches... 99 00:06:41,080 --> 00:06:42,560 ..and it worked. 100 00:06:43,800 --> 00:06:46,560 They're now in 155 countries 101 00:06:46,560 --> 00:06:48,120 with offices across the globe. 102 00:06:49,400 --> 00:06:52,440 Nearly a third of all humans are connected. 103 00:06:53,680 --> 00:06:55,280 So if you look at this you might think 104 00:06:55,280 --> 00:06:58,400 this is actually a map of the world but it's not. 105 00:06:58,400 --> 00:07:01,200 It's a blank screen and then, for every person who's using Facebook, 106 00:07:01,200 --> 00:07:03,640 we put a dot on the screen, 107 00:07:03,640 --> 00:07:06,280 and then where are their friend connections and we draw lines. 108 00:07:06,280 --> 00:07:07,680 And when you do that, 109 00:07:07,680 --> 00:07:09,360 it effectively draws the world. 110 00:07:11,160 --> 00:07:13,360 I think the growth has really surprised everybody. 111 00:07:13,360 --> 00:07:14,400 It's like, "Wow." 112 00:07:16,120 --> 00:07:19,680 By tapping into core human needs to share, be heard, 113 00:07:19,680 --> 00:07:22,080 and witness the lives of others, 114 00:07:22,080 --> 00:07:24,760 they've grown a business in just 15 years 115 00:07:24,760 --> 00:07:26,520 worth half $1 trillion. 116 00:07:29,840 --> 00:07:31,360 For the people who work here, 117 00:07:31,360 --> 00:07:33,960 it's more than just a job. 118 00:07:33,960 --> 00:07:36,280 It's a mission. 119 00:07:36,280 --> 00:07:41,200 I was really driven here by the mission and vision of connectivity, 120 00:07:41,200 --> 00:07:45,280 of getting everyone around the world democratised access to information. 121 00:07:45,280 --> 00:07:50,120 We believe it's our responsibility to bring the world together 122 00:07:50,120 --> 00:07:54,000 and to give power to the community to build a better world. 123 00:07:55,360 --> 00:07:59,360 This is a business that rewrote the rules for how a company should be. 124 00:08:00,800 --> 00:08:04,160 Employees are encouraged to play, experiment, 125 00:08:04,160 --> 00:08:06,040 and think outside the box. 126 00:08:07,720 --> 00:08:10,440 You know, there are other companies, like Apple, for example, 127 00:08:10,440 --> 00:08:12,720 who you know, their products need to be perfect 128 00:08:12,720 --> 00:08:14,000 before they launch it. 129 00:08:14,000 --> 00:08:15,760 Whereas we want to launch something quickly 130 00:08:15,760 --> 00:08:18,800 and then continuously improve it. 131 00:08:18,800 --> 00:08:22,120 You have to encourage people to take risks and be willing to fail 132 00:08:22,120 --> 00:08:24,800 because otherwise you're just not going to achieve anything. 133 00:08:27,120 --> 00:08:29,800 This successful risk-taking strategy 134 00:08:29,800 --> 00:08:32,240 was embodied by their early mantra... 135 00:08:37,320 --> 00:08:41,640 But as 2018 got under way, it became clear things had become 136 00:08:41,640 --> 00:08:43,000 very broken indeed. 137 00:08:46,080 --> 00:08:49,640 Information from millions of Facebook users may have been used 138 00:08:49,640 --> 00:08:53,920 by a data company during the 2016 US presidential election. 139 00:08:53,920 --> 00:08:57,200 A former employee of Cambridge Analytica claims that 140 00:08:57,200 --> 00:08:59,200 50 million profiles were accessed. 141 00:08:59,200 --> 00:09:02,680 The biggest data breach in Facebook history. 142 00:09:05,560 --> 00:09:10,000 In March 2018, the first major scandal to rock Facebook broke. 143 00:09:12,240 --> 00:09:14,840 An external app developer exploited a loophole, 144 00:09:14,840 --> 00:09:17,320 enabling them to harvest the data of 145 00:09:17,320 --> 00:09:20,760 87 million Facebook users as well as their friends. 146 00:09:22,840 --> 00:09:27,680 This was sold to political consultancy Cambridge Analytica. 147 00:09:27,680 --> 00:09:30,800 They allegedly used it to help political organisations 148 00:09:30,800 --> 00:09:33,800 influence voters on Facebook. 149 00:09:33,800 --> 00:09:36,880 Facebook faces the wrath of regulators and customers 150 00:09:36,880 --> 00:09:38,400 in multiple countries. 151 00:09:38,400 --> 00:09:41,480 Its reputation has sunk, along with its share price, 152 00:09:41,480 --> 00:09:46,520 and a business plan based on liberal sharing of data may be under threat. 153 00:09:46,520 --> 00:09:48,920 The crisis deepened when it emerged 154 00:09:48,920 --> 00:09:52,200 Russia had also used Facebook to push fake news. 155 00:09:55,480 --> 00:09:57,920 Facebook's interference in our elections 156 00:09:57,920 --> 00:09:59,440 and in our society 157 00:09:59,440 --> 00:10:01,360 has de-stabled our democracies. 158 00:10:03,000 --> 00:10:04,240 To make matters worse, 159 00:10:04,240 --> 00:10:07,680 the platform had become a breeding ground for hate speech 160 00:10:07,680 --> 00:10:09,320 and exploitative scams. 161 00:10:11,680 --> 00:10:15,800 The very connectivity and openness Facebook had built its success on 162 00:10:15,800 --> 00:10:18,440 risked becoming its greatest weakness. 163 00:10:21,880 --> 00:10:24,320 Join the trend of deleting Facebook altogether. 164 00:10:24,320 --> 00:10:25,400 #deletefacebook. 165 00:10:36,320 --> 00:10:41,040 Mark Zuckerberg was called before Congress to testify. 166 00:10:41,040 --> 00:10:44,320 I think it's time to ask whether Facebook may have moved too fast 167 00:10:44,320 --> 00:10:46,440 and broken too many things. 168 00:10:46,440 --> 00:10:47,680 I think the damage done 169 00:10:47,680 --> 00:10:50,800 to our democracy, relative to 170 00:10:50,800 --> 00:10:52,680 Facebook and its platform 171 00:10:52,680 --> 00:10:55,400 being weaponised, are incalculable. 172 00:10:56,440 --> 00:11:00,200 Facebook has grown so big, so fast. 173 00:11:00,200 --> 00:11:03,160 That much influence comes with 174 00:11:03,160 --> 00:11:06,960 enormous social responsibility 175 00:11:06,960 --> 00:11:09,200 of which you have failed to act. 176 00:11:12,480 --> 00:11:13,880 For most of our existence, 177 00:11:13,880 --> 00:11:15,520 we focused on all the good 178 00:11:15,520 --> 00:11:17,760 that connecting people can bring. 179 00:11:17,760 --> 00:11:20,920 But it's clear now that we didn't do enough 180 00:11:20,920 --> 00:11:24,440 to prevent these tools from being used for harm as well. 181 00:11:24,440 --> 00:11:28,480 And that goes for fake news, foreign interference in elections, 182 00:11:28,480 --> 00:11:29,920 and hate speech, 183 00:11:29,920 --> 00:11:33,400 as well as developers and data privacy. 184 00:11:33,400 --> 00:11:37,240 We didn't take a broad enough view of our responsibility. 185 00:11:37,240 --> 00:11:39,480 And that was a big mistake. 186 00:11:39,480 --> 00:11:42,080 It was my mistake and I'm sorry. 187 00:12:00,600 --> 00:12:04,600 Following the crisis, it was clear that, inside Facebook, 188 00:12:04,600 --> 00:12:06,480 big changes had to be made. 189 00:12:08,920 --> 00:12:10,560 The aftermath of Cambridge Analytica 190 00:12:10,560 --> 00:12:12,800 was very painful time for us internally. 191 00:12:14,200 --> 00:12:16,440 We all deeply care about what we're doing here. 192 00:12:16,440 --> 00:12:19,720 We all want to feel we're doing something positive in the world. 193 00:12:20,920 --> 00:12:24,440 That, you know...that faith was a little bit crushed. 194 00:12:24,440 --> 00:12:26,840 And so that was really painful for a lot of people. 195 00:12:30,480 --> 00:12:34,200 Learning, you know, from... 196 00:12:34,200 --> 00:12:36,480 ..mistakes we've made in the past. 197 00:12:37,560 --> 00:12:39,400 Let's look forward. 198 00:12:39,400 --> 00:12:41,640 You know, where from here. 199 00:12:41,640 --> 00:12:46,120 How do we build in order to make sure that people can trust us? 200 00:12:48,680 --> 00:12:51,400 I just hope we have enough of a measure to conversation around it, 201 00:12:51,400 --> 00:12:54,880 to not throw out what is a lot of powerful and good opportunity 202 00:12:54,880 --> 00:12:57,000 that these tools create... 203 00:12:57,000 --> 00:12:59,360 ..in the pursuit of trying to limit the damage 204 00:12:59,360 --> 00:13:01,480 they create alongside. 205 00:13:10,960 --> 00:13:14,240 Facebook have fixed the loophole that enabled Cambridge Analytica 206 00:13:14,240 --> 00:13:16,600 to harvest personal data. 207 00:13:16,600 --> 00:13:19,320 But this scandal was just the beginning. 208 00:13:21,720 --> 00:13:27,200 Hate speech, fake news and scams continue to affect public trust 209 00:13:27,200 --> 00:13:29,800 and keep them in the media spotlight. 210 00:13:32,920 --> 00:13:36,080 In response, Facebook has hired 30,000 more people 211 00:13:36,080 --> 00:13:38,120 to help police the platform. 212 00:13:40,040 --> 00:13:43,400 The question is - can you fix a problem so huge? 213 00:14:04,480 --> 00:14:07,760 Before I came to Facebook, I was working as a criminal prosecutor 214 00:14:07,760 --> 00:14:09,760 for the United States Justice Department, 215 00:14:09,760 --> 00:14:12,800 and I worked on things like stopping terrorism, 216 00:14:12,800 --> 00:14:16,280 responding to child abuse, human trafficking. 217 00:14:17,720 --> 00:14:19,960 But one thing I've learned from working at Facebook 218 00:14:19,960 --> 00:14:22,400 is the same bad behaviour that you see offline, 219 00:14:22,400 --> 00:14:23,600 you're going to see online. 220 00:14:23,600 --> 00:14:25,680 And so, it's about making sure 221 00:14:25,680 --> 00:14:28,720 that we are creating the space for all the really great people 222 00:14:28,720 --> 00:14:30,840 on the site who are coming for great reasons, 223 00:14:30,840 --> 00:14:32,960 but also putting up some barriers that the bad actors 224 00:14:32,960 --> 00:14:34,160 are not a part of the community. 225 00:14:36,400 --> 00:14:39,120 Monica's been tasked with setting the rules 226 00:14:39,120 --> 00:14:42,280 that will define what people can and cannot post. 227 00:14:42,280 --> 00:14:45,720 With the aim of blocking abuse like hate speech. 228 00:14:47,560 --> 00:14:50,280 In Germany, violence against refugees 229 00:14:50,280 --> 00:14:52,920 has been linked to posts inciting extremism. 230 00:14:55,160 --> 00:14:58,960 In Myanmar, Facebook has been used to foment hatred 231 00:14:58,960 --> 00:15:01,840 against the Rohingya and other Muslims. 232 00:15:01,840 --> 00:15:05,960 The UN has accused the platform of contributing to ethnic cleansing. 233 00:15:08,760 --> 00:15:12,360 But with Facebook in over 155 countries, 234 00:15:12,360 --> 00:15:15,200 how do you write rules allowing for freedom of speech 235 00:15:15,200 --> 00:15:17,840 across every culture and religion... 236 00:15:19,800 --> 00:15:22,400 ..whilst also stopping abuse? 237 00:15:27,000 --> 00:15:29,760 This is representative of the kind of content we're seeing 238 00:15:29,760 --> 00:15:31,480 at the moment in the UK. 239 00:15:31,480 --> 00:15:33,880 We have this image of a police officer, 240 00:15:33,880 --> 00:15:36,960 and she's talking to a woman in a burqa holding a child, 241 00:15:36,960 --> 00:15:40,760 and the text overlay here says, "Yeah, turn left, walk one mile, 242 00:15:40,760 --> 00:15:43,040 "then piss off back to your own country." 243 00:15:44,680 --> 00:15:47,600 We're going off the burqa as an indicator that 244 00:15:47,600 --> 00:15:51,160 the woman is being attacked for being Muslim. 245 00:15:51,160 --> 00:15:53,200 We wondered about how we would treat this 246 00:15:53,200 --> 00:15:55,200 if the burqa was not present, 247 00:15:55,200 --> 00:15:57,800 what the policy view would be on this. 248 00:15:57,800 --> 00:16:02,360 Or if this was appearing somewhere where, you know, 249 00:16:02,360 --> 00:16:05,160 this was critical commentary about law enforcement 250 00:16:05,160 --> 00:16:10,120 and how they are not kind to immigrants who are Muslims. 251 00:16:11,720 --> 00:16:14,760 When I came to Facebook, I had been living overseas 252 00:16:14,760 --> 00:16:16,840 and dealing with international law for some time, 253 00:16:16,840 --> 00:16:19,040 so I understood that people around the world would have 254 00:16:19,040 --> 00:16:22,040 really different ideas about where our policy should be, 255 00:16:22,040 --> 00:16:24,280 but one thing that I didn't anticipate was 256 00:16:24,280 --> 00:16:29,040 how difficult it is to craft rules that work at our size. 257 00:16:29,040 --> 00:16:32,080 With the volume of content we see every day, 258 00:16:32,080 --> 00:16:35,040 we get millions of reports every week, 259 00:16:35,040 --> 00:16:38,800 and so we need to craft rules that can be enforced fairly 260 00:16:38,800 --> 00:16:42,480 and efficiently and quickly around the world. 261 00:16:42,480 --> 00:16:46,000 Facebook has seen a huge rise in hate speech, 262 00:16:46,000 --> 00:16:48,840 and with the speed language is evolving online, 263 00:16:48,840 --> 00:16:51,080 they're constantly playing catch-up. 264 00:16:52,320 --> 00:16:55,960 It's not always obvious if something is hate speech or not. 265 00:16:57,320 --> 00:17:00,440 In this particular instance, you can see that there's no text. 266 00:17:00,440 --> 00:17:02,240 We just have emojis. 267 00:17:02,240 --> 00:17:06,040 So, this is two men holding hands and then a gun. 268 00:17:06,040 --> 00:17:10,000 We do know that these men are often symbolising a gay couple, 269 00:17:10,000 --> 00:17:12,840 so it could be an attack on the protected characteristic. 270 00:17:12,840 --> 00:17:15,840 Then, obviously, with the gun here, that's a threat of violence, 271 00:17:15,840 --> 00:17:18,920 so we will be taking this as an attack on gay people 272 00:17:18,920 --> 00:17:20,520 and removing for hate speech. 273 00:17:20,520 --> 00:17:23,120 Basically, more and more, we are seeing people 274 00:17:23,120 --> 00:17:26,720 using emojis on the platform to symbolise hate. 275 00:17:26,720 --> 00:17:28,640 I think it's a good example of how, like, 276 00:17:28,640 --> 00:17:31,920 users are getting more creative to bypass our policies. 277 00:17:31,920 --> 00:17:35,160 Yeah. Just don't put it in words. Put it in pictures or emojis. Yes. 278 00:17:35,160 --> 00:17:36,600 Yeah. Absolutely. 279 00:17:36,600 --> 00:17:42,280 There are no templates, really, for doing this kind of work. 280 00:17:42,280 --> 00:17:44,080 Our job is to try to get ahead of it, 281 00:17:44,080 --> 00:17:46,320 and we do that by talking to experts, 282 00:17:46,320 --> 00:17:48,680 talking to organisations around the world, 283 00:17:48,680 --> 00:17:51,560 hearing about trends that are happening in the offline world, 284 00:17:51,560 --> 00:17:53,240 even before they come online, 285 00:17:53,240 --> 00:17:56,320 and trying to put rules in place to address them. 286 00:17:56,320 --> 00:17:58,720 This horrific word in a kind context... 287 00:17:58,720 --> 00:18:01,960 Monica's rules are based on US federal law, 288 00:18:01,960 --> 00:18:06,840 which protects characteristics like race, religion or gender, 289 00:18:06,840 --> 00:18:09,320 but these rules also need enforcing. 290 00:18:12,760 --> 00:18:16,800 Every region has its own team of content reviewers 291 00:18:16,800 --> 00:18:21,080 who use the rules to decide whether a post should be removed or not. 292 00:18:22,880 --> 00:18:25,760 Millions of posts are taken down every week, 293 00:18:25,760 --> 00:18:28,160 but it's a challenge that is only growing. 294 00:18:30,640 --> 00:18:34,320 We want Facebook to be a place where people want to come, 295 00:18:34,320 --> 00:18:37,920 and we know that they aren't going to do that if it's not a safe place. 296 00:18:39,560 --> 00:18:41,400 At our scale, at our volume, 297 00:18:41,400 --> 00:18:45,160 even if we get it right 99% of the time, 298 00:18:45,160 --> 00:18:48,000 that still means we're making thousands of mistakes every day, 299 00:18:48,000 --> 00:18:51,040 and every one of those mistakes is important to somebody. 300 00:18:53,360 --> 00:18:55,360 Facebook have no way of knowing 301 00:18:55,360 --> 00:18:57,960 how much hate speech is on their platform. 302 00:18:59,680 --> 00:19:03,000 Zuckerberg has promised thousands more reviewers 303 00:19:03,000 --> 00:19:06,640 but, with billions of posts and comments every day, 304 00:19:06,640 --> 00:19:10,240 it's now too big to police with manpower alone. 305 00:19:13,000 --> 00:19:17,120 Without technological innovation, they're in danger of losing control. 306 00:19:21,280 --> 00:19:24,640 TRAM BELL DINGS 307 00:19:37,240 --> 00:19:38,960 Bonjour. Bonjour. 308 00:19:38,960 --> 00:19:41,600 Mark said, you know, "How can you help us?" 309 00:19:41,600 --> 00:19:44,400 He invited me to dinner at his house and we had a long chat 310 00:19:44,400 --> 00:19:48,200 on what his vision was for AI and what my vision was. 311 00:19:48,200 --> 00:19:50,040 When someone tells you, you know, 312 00:19:50,040 --> 00:19:52,480 "Come to us and start a research lab from scratch," 313 00:19:52,480 --> 00:19:54,720 you know, it's a very interesting challenge. 314 00:19:56,000 --> 00:19:59,560 Yann is a world-leading expert in artificial intelligence. 315 00:20:00,960 --> 00:20:04,640 He's building Facebook's multi-million-dollar AI lab. 316 00:20:06,680 --> 00:20:09,760 Currently, their AI can detect faces 317 00:20:09,760 --> 00:20:12,080 and automatically take down clear abuse, 318 00:20:12,080 --> 00:20:14,960 like pornography and terrorist content... 319 00:20:16,840 --> 00:20:20,520 ..but what's harder is getting a machine to understand the context 320 00:20:20,520 --> 00:20:23,560 and meaning of a post the way a human can. 321 00:20:26,160 --> 00:20:28,600 It's a huge technological challenge, 322 00:20:28,600 --> 00:20:32,560 but vital if Facebook is to tackle the scale of the abuse. 323 00:20:34,760 --> 00:20:38,280 What the internal systems of Facebook have to do is basically 324 00:20:38,280 --> 00:20:44,200 understand text in any language, images, video, speech, whatever, 325 00:20:44,200 --> 00:20:45,680 but it's difficult. 326 00:20:45,680 --> 00:20:50,000 What if it's, you know, the iconic picture 327 00:20:50,000 --> 00:20:53,080 of the little girl running away from napalm bombing in Vietnam, right? 328 00:20:53,080 --> 00:20:57,040 That's an iconic, journalistic piece that, you know, 329 00:20:57,040 --> 00:20:59,320 has historical value, 330 00:20:59,320 --> 00:21:01,480 but, on the face of it, it's child pornography. 331 00:21:02,720 --> 00:21:04,880 There's always a certain percentage of mistakes, 332 00:21:04,880 --> 00:21:06,360 and so the question is, you know, 333 00:21:06,360 --> 00:21:09,080 how to design a system to minimise those errors 334 00:21:09,080 --> 00:21:11,680 without upsetting too many people, essentially. 335 00:21:11,680 --> 00:21:14,960 So, the technology for this is not simple. It's not perfect yet. 336 00:21:17,240 --> 00:21:20,240 Next door, one of Yann's researchers 337 00:21:20,240 --> 00:21:23,400 is helping to solve some of these problems in videos. 338 00:21:23,400 --> 00:21:25,280 OK, there you go. 339 00:21:25,280 --> 00:21:27,560 That's the thing I wanted to show you. 340 00:21:27,560 --> 00:21:32,400 Iasonas has built some AI that can read human movement - 341 00:21:32,400 --> 00:21:37,080 a stepping stone to getting computers to understand context. 342 00:21:37,080 --> 00:21:38,920 OK, so our system never 343 00:21:38,920 --> 00:21:41,160 saw this video before. 344 00:21:41,160 --> 00:21:45,200 This was a video we gathered after our system had been trained, 345 00:21:45,200 --> 00:21:48,920 and this is actually, on the fly, identifying, 346 00:21:48,920 --> 00:21:51,080 for every pixel that belongs to a human, 347 00:21:51,080 --> 00:21:54,320 where it lands on the body's surface. 348 00:21:54,320 --> 00:21:57,440 So, if I, let's say, move my arm around like this, 349 00:21:57,440 --> 00:22:00,320 my joint position stays exactly the same, OK? 350 00:22:00,320 --> 00:22:03,880 So, a landmark-based system is blind to this motion 351 00:22:03,880 --> 00:22:05,280 while, with our system, 352 00:22:05,280 --> 00:22:07,760 we can actually track the surface as it's moving around. 353 00:22:09,200 --> 00:22:11,800 The objective is to be able to understand 354 00:22:11,800 --> 00:22:13,640 what humans are doing in videos 355 00:22:13,640 --> 00:22:17,280 so that you can understand their body language, their emotions, 356 00:22:17,280 --> 00:22:19,360 what they want to do, and so on. 357 00:22:19,360 --> 00:22:21,960 Of course, there are many direct connections 358 00:22:21,960 --> 00:22:23,600 to the needs of Facebook - 359 00:22:23,600 --> 00:22:26,600 understanding whether a post involves violence, 360 00:22:26,600 --> 00:22:28,840 terrorism, propaganda or whatever. 361 00:22:28,840 --> 00:22:30,720 So, you want to be able to understand 362 00:22:30,720 --> 00:22:32,360 what humans are doing in images. 363 00:22:34,360 --> 00:22:38,560 The vision of AI long-term is for Facebook servers 364 00:22:38,560 --> 00:22:41,800 to understand the meaning and context behind every post 365 00:22:41,800 --> 00:22:43,720 and comment in real time... 366 00:22:44,760 --> 00:22:47,360 ..spotting violence, abusive content 367 00:22:47,360 --> 00:22:50,720 and even fake news before it can be shared 368 00:22:50,720 --> 00:22:53,680 without the need for human review. 369 00:22:53,680 --> 00:22:57,160 We see a huge advantage from technology. 370 00:22:57,160 --> 00:23:02,400 Already, there are a small number of areas where the technology 371 00:23:02,400 --> 00:23:04,440 can look at something that's on the site and say, 372 00:23:04,440 --> 00:23:06,440 "We know this violates our policies." 373 00:23:06,440 --> 00:23:08,680 And in those small set of circumstances, 374 00:23:08,680 --> 00:23:12,080 that content doesn't need to go to people to make that decision. 375 00:23:12,080 --> 00:23:15,600 And as time goes by, that technology is getting better and better, 376 00:23:15,600 --> 00:23:17,960 and I think we'll continue to use it more and more. 377 00:23:19,240 --> 00:23:21,680 This technology could be the thing 378 00:23:21,680 --> 00:23:24,480 that will prevent abusive posts in the future. 379 00:23:25,640 --> 00:23:28,160 But if their AI is going to be smart enough 380 00:23:28,160 --> 00:23:31,640 to do the job of thousands of content reviewers, 381 00:23:31,640 --> 00:23:34,080 Facebook needs their network 382 00:23:34,080 --> 00:23:36,800 to have the kind of intelligence we do. 383 00:23:38,160 --> 00:23:41,920 I'd like to find kind of an organising principle 384 00:23:41,920 --> 00:23:44,040 that would allow us to get machines to learn 385 00:23:44,040 --> 00:23:47,080 a little bit like animals and humans - by observation. 386 00:23:48,320 --> 00:23:51,760 As to the question of whether it is possible to build machines 387 00:23:51,760 --> 00:23:55,000 that have kind of the same level of intelligence as humans 388 00:23:55,000 --> 00:23:56,520 in kind of all respects, 389 00:23:56,520 --> 00:23:59,400 it's very difficult to predict how long that's going to take, 390 00:23:59,400 --> 00:24:02,920 but that's pretty much a matter of time. 391 00:24:08,360 --> 00:24:10,560 While Facebook have been working on a rule book 392 00:24:10,560 --> 00:24:14,400 and technology to remove abusive content, 393 00:24:14,400 --> 00:24:19,200 it's also become an effective weapon for organised criminals. 394 00:24:19,200 --> 00:24:23,040 They call it "sextortion". Really, it's online blackmail. 395 00:24:23,040 --> 00:24:25,880 Scammers can make hundreds of dollars a day, 396 00:24:25,880 --> 00:24:30,160 and this crime is now happening on an industrial scale. 397 00:24:35,240 --> 00:24:37,960 Criminals are using Facebook's global reach 398 00:24:37,960 --> 00:24:40,960 to target victims with devastating effect... 399 00:24:42,600 --> 00:24:47,000 ..from human trafficking to child abuse and financial scams. 400 00:24:51,440 --> 00:24:53,920 It's become such an immense challenge, 401 00:24:53,920 --> 00:24:57,360 Facebook have built a new department to try to confront it. 402 00:25:20,760 --> 00:25:23,800 In London, a team of specialists has been assembled 403 00:25:23,800 --> 00:25:27,360 to target these types of scams and criminal behaviour. 404 00:25:29,880 --> 00:25:33,120 We feel a lot more pressure to go faster. 405 00:25:34,320 --> 00:25:38,240 You know, we take it personally. Some of the stuff is... 406 00:25:38,240 --> 00:25:40,440 We wish we would have done more 407 00:25:40,440 --> 00:25:42,480 or we see that we could have done... 408 00:25:42,480 --> 00:25:45,520 We made a mistake here and there, so, you know, 409 00:25:45,520 --> 00:25:48,240 we view this as our core responsibility - 410 00:25:48,240 --> 00:25:49,560 to fix this. 411 00:25:52,680 --> 00:25:54,760 Organised criminals are always looking 412 00:25:54,760 --> 00:25:56,880 for new avenues to exploit... 413 00:25:59,120 --> 00:26:01,560 ..and abuse like financial scams on Facebook 414 00:26:01,560 --> 00:26:04,000 are taking many different forms. 415 00:26:10,280 --> 00:26:13,160 There's people who, like, claim to have ins on the betting industry, 416 00:26:13,160 --> 00:26:14,440 so they'll tell people, like, 417 00:26:14,440 --> 00:26:16,280 they have a guaranteed way to make money. 418 00:26:16,280 --> 00:26:18,560 You see people selling, like, fake cryptocurrencies, 419 00:26:18,560 --> 00:26:20,000 people selling, like, puppies, 420 00:26:20,000 --> 00:26:22,320 and then, you know, extorting people for money on them. 421 00:26:22,320 --> 00:26:23,960 People are pretty creative and, yeah, 422 00:26:23,960 --> 00:26:26,520 they're constantly innovating, which keeps it challenging. 423 00:26:27,960 --> 00:26:30,840 Stopping this serious abuse is made harder 424 00:26:30,840 --> 00:26:35,320 because the majority actually come from fake Facebook accounts. 425 00:26:35,320 --> 00:26:38,280 These are made to look real so users trust them, 426 00:26:38,280 --> 00:26:41,760 but they're also incredibly hard to trace, 427 00:26:41,760 --> 00:26:44,600 and they're a growing problem. 428 00:26:44,600 --> 00:26:48,800 Every day, millions now appear on the network, 429 00:26:48,800 --> 00:26:52,520 and most of them are created automatically 430 00:26:52,520 --> 00:26:54,760 using computer programs. 431 00:26:56,200 --> 00:26:59,320 These fake accounts are being made all over the world. 432 00:26:59,320 --> 00:27:01,640 You know, there are lots and lots of ways 433 00:27:01,640 --> 00:27:03,520 by which people create them. 434 00:27:03,520 --> 00:27:06,160 Often, they hide their tracks. 435 00:27:06,160 --> 00:27:09,280 Somebody hijacks hundreds of thousands of machines, 436 00:27:09,280 --> 00:27:11,040 like the old computer-style virus, 437 00:27:11,040 --> 00:27:14,080 and they use these machines all over the world 438 00:27:14,080 --> 00:27:16,320 to act on their behalf. 439 00:27:18,040 --> 00:27:22,360 If they can stop fake accounts, they can stop scams. 440 00:27:22,360 --> 00:27:25,400 In the early days, Facebook largely relied on users 441 00:27:25,400 --> 00:27:27,600 to report suspicious accounts, 442 00:27:27,600 --> 00:27:30,200 leaving a lot of fake ones unchecked. 443 00:27:32,520 --> 00:27:33,840 In the past year, 444 00:27:33,840 --> 00:27:36,960 they've managed to develop artificial intelligence software 445 00:27:36,960 --> 00:27:41,040 to analyse the behaviour of all 2.3 billion accounts 446 00:27:41,040 --> 00:27:43,760 and spot the fake ones automatically. 447 00:27:47,000 --> 00:27:49,200 Behaviours get reflected within the system, 448 00:27:49,200 --> 00:27:52,080 you know, for example, like the geography that you're coming from, 449 00:27:52,080 --> 00:27:53,760 how you connect to the internet - 450 00:27:53,760 --> 00:27:57,000 all sorts of information that might not be obvious 451 00:27:57,000 --> 00:28:00,440 to a human being, but it makes a ton of sense to a computer. 452 00:28:00,440 --> 00:28:03,280 I'll give you a very simplified example. 453 00:28:03,280 --> 00:28:06,800 You know, most people have, you know, a relationship, 454 00:28:06,800 --> 00:28:10,120 maybe multiple relationships at the same time. 455 00:28:10,120 --> 00:28:12,800 You're not likely to have 456 00:28:12,800 --> 00:28:18,360 50 online romantic relationships all going at once at various stages, 457 00:28:18,360 --> 00:28:20,240 so that looks suspicious. 458 00:28:21,320 --> 00:28:24,960 Their new AI has been observing the way people interact, 459 00:28:24,960 --> 00:28:28,200 looking for anomalies in account history. 460 00:28:28,200 --> 00:28:30,040 Since the beginning of the year, 461 00:28:30,040 --> 00:28:32,720 it's taken down more than a billion fake accounts. 462 00:28:34,240 --> 00:28:36,880 It's the most powerful defence they have, 463 00:28:36,880 --> 00:28:39,400 so filming the code is not allowed. 464 00:28:40,800 --> 00:28:43,800 INTERVIEWER: Where essentially is the artificial intelligence? 465 00:28:43,800 --> 00:28:46,920 We don't publish the exact techniques that we use. 466 00:28:46,920 --> 00:28:48,520 It's... 467 00:28:48,520 --> 00:28:51,160 The artificial intelligence, or the machine learning, 468 00:28:51,160 --> 00:28:55,200 is running actually in data centres in various places. 469 00:28:55,200 --> 00:28:58,720 What you're seeing here are people who are building programs 470 00:28:58,720 --> 00:29:02,960 that collect the data needed to actually make these things work. 471 00:29:02,960 --> 00:29:05,920 In spite of this, more appear every day, 472 00:29:05,920 --> 00:29:07,840 because the people on the other side 473 00:29:07,840 --> 00:29:10,640 keep finding new ways to cheat their AI. 474 00:29:12,720 --> 00:29:15,760 When something starts working, for example, a detection technique, 475 00:29:15,760 --> 00:29:18,400 we see people going, "Let's try this a little bit." 476 00:29:18,400 --> 00:29:21,040 These guys actually poke and prod 477 00:29:21,040 --> 00:29:24,960 and test defences actively all the time. 478 00:29:24,960 --> 00:29:27,520 You know, every once in a while, they'll find a hole, 479 00:29:27,520 --> 00:29:30,360 and we need to make sure we detect when they find a hole, 480 00:29:30,360 --> 00:29:33,280 and plug it, and we keep moving forward like that. 481 00:29:37,560 --> 00:29:41,080 Facebook is caught in a new kind of cyber war. 482 00:29:42,760 --> 00:29:45,240 All of them are getting more and more advanced. 483 00:29:45,240 --> 00:29:47,520 People are pouring resources and money into it, 484 00:29:47,520 --> 00:29:49,200 and the more money they can make, 485 00:29:49,200 --> 00:29:51,680 the more money they are going to be willing to spend 486 00:29:51,680 --> 00:29:53,960 to kind of break your systems. 487 00:29:56,560 --> 00:29:58,680 What is your greatest threat? 488 00:29:59,840 --> 00:30:01,760 What is my greatest threat? 489 00:30:04,400 --> 00:30:06,360 Staying on top of it. 490 00:30:06,360 --> 00:30:09,360 That's like, if we don't stay on top of it, 491 00:30:09,360 --> 00:30:12,000 we know, like, if we just stand still, 492 00:30:12,000 --> 00:30:13,640 we know we will lose. 493 00:30:15,360 --> 00:30:17,720 So, it's kind of like... 494 00:30:17,720 --> 00:30:19,520 Anyway, that's it. 495 00:30:41,520 --> 00:30:43,320 By the summer of 2018, 496 00:30:43,320 --> 00:30:45,520 Facebook's problems were deepening... 497 00:30:47,360 --> 00:30:49,800 ..from failure to protect personal data 498 00:30:49,800 --> 00:30:52,440 to inciting hate crime and scams. 499 00:30:54,280 --> 00:30:56,400 It all came to a head in July. 500 00:30:57,960 --> 00:31:01,880 Facebook is to be fined £500,000 by the Information Commissioner. 501 00:31:01,880 --> 00:31:05,520 That's the maximum fine possible for misuse of data in the UK. 502 00:31:05,520 --> 00:31:07,680 Facebook is going to have to cough up 503 00:31:07,680 --> 00:31:10,120 because it failed to protect users' information 504 00:31:10,120 --> 00:31:12,680 from the third-party firm Cambridge Analytica. 505 00:31:12,680 --> 00:31:14,560 The proposed fine for Facebook 506 00:31:14,560 --> 00:31:17,240 may be small change for an online giant, 507 00:31:17,240 --> 00:31:19,960 but it's still a powerful statement of intent, 508 00:31:19,960 --> 00:31:22,840 and in future, fines could be much bigger. 509 00:31:28,400 --> 00:31:32,240 In spite of all the scandals, half a million more people 510 00:31:32,240 --> 00:31:35,680 are still signing up to Facebook every single day. 511 00:31:48,320 --> 00:31:50,960 Despite the negative rolling news coverage, 512 00:31:50,960 --> 00:31:52,880 inside Facebook, 513 00:31:52,880 --> 00:31:56,040 the business of keeping the platform running goes on. 514 00:31:59,080 --> 00:32:01,560 The staff are continuing to move fast 515 00:32:01,560 --> 00:32:04,680 and embrace the unique hack culture that's got them this far. 516 00:32:16,520 --> 00:32:19,240 This is an organisation that rewards risk 517 00:32:19,240 --> 00:32:20,800 and embraces failure. 518 00:32:27,480 --> 00:32:29,760 Every few months, each office get together 519 00:32:29,760 --> 00:32:32,600 for their own 24-hour lock-in 520 00:32:32,600 --> 00:32:37,160 where staff are encouraged to innovate and try new ideas. 521 00:32:37,160 --> 00:32:39,840 It's what they call a "hackathon" 522 00:32:39,840 --> 00:32:43,360 and it's how they came up with the "like" button. 523 00:32:43,360 --> 00:32:45,200 Coding? It's like cooking, big man. 524 00:32:45,200 --> 00:32:48,040 You work for hours to find the perfect recipe, 525 00:32:48,040 --> 00:32:51,560 and once you've found it, bingo - you sell it to the world. 526 00:32:56,360 --> 00:32:59,800 At Facebook, all employees have two birthdays - 527 00:32:59,800 --> 00:33:02,440 their own and their Faceversary, 528 00:33:02,440 --> 00:33:05,680 which they celebrate each year they're at the company. 529 00:33:07,200 --> 00:33:09,960 They have Bring Your Family To Work Days, 530 00:33:09,960 --> 00:33:11,400 Pride events, 531 00:33:11,400 --> 00:33:14,560 and they get celebrities in to inspire the workforce. 532 00:33:18,680 --> 00:33:23,760 They even have a small village with bikes, shops and an arcade. 533 00:33:23,760 --> 00:33:29,120 Everything is free, so everyone can focus on connecting the world. 534 00:33:32,520 --> 00:33:35,960 To help fix the scale of the problems that have been exposed, 535 00:33:35,960 --> 00:33:40,400 they're hiring more engineers at a staggering rate. 536 00:33:40,400 --> 00:33:42,760 These are the people who will be responsible 537 00:33:42,760 --> 00:33:45,280 for taking Facebook into the future. 538 00:33:46,440 --> 00:33:50,400 Today, they're being initiated into the unique culture. 539 00:33:50,400 --> 00:33:53,240 If you haven't already, on the left-hand side of the room, 540 00:33:53,240 --> 00:33:55,040 you can fill in a name badge. 541 00:33:55,040 --> 00:33:58,920 In the last year, they've hired more than 10,000 people. 542 00:34:00,120 --> 00:34:04,000 Every week, 50 new engineers arrive in London. 543 00:34:04,000 --> 00:34:06,640 Hello, everyone. This is Ray Ming from China, 544 00:34:06,640 --> 00:34:09,880 and I will be joining the London office as a software engineer. 545 00:34:09,880 --> 00:34:14,560 Hi, my name's Marina, joining in London. Originally Russian. 546 00:34:14,560 --> 00:34:17,760 With the company set to double in the next few years, 547 00:34:17,760 --> 00:34:21,640 it's these engineers they're now relying on to fix things, 548 00:34:21,640 --> 00:34:23,680 as well as grow the business. 549 00:34:24,960 --> 00:34:27,520 When we say this is now your company, we really mean it. 550 00:34:27,520 --> 00:34:30,400 This isn't just any other company. This is now YOUR company. 551 00:34:30,400 --> 00:34:33,960 So, it's in your hands to make this company better or worse. 552 00:34:35,320 --> 00:34:38,240 When we give feedback, we give feedback in a way that is constructive 553 00:34:38,240 --> 00:34:41,560 and it's not just showing that you are the smartest person in the room. 554 00:34:41,560 --> 00:34:43,600 We do talk about feedback being a gift, 555 00:34:43,600 --> 00:34:46,480 but we don't believe in being arseholes, and that's like... 556 00:34:46,480 --> 00:34:48,880 It's straight up. We just... We don't tolerate dicks. 557 00:34:49,920 --> 00:34:53,120 Everything we do, we're doing to connect people, 558 00:34:53,120 --> 00:34:57,000 and we're doing to try and bring social value in that connection. 559 00:34:57,000 --> 00:34:59,640 That means that every product decision that we make 560 00:34:59,640 --> 00:35:03,720 is tuned for, "Does it do good in the world?" 561 00:35:03,720 --> 00:35:06,360 We're not a company that's designed to make money. 562 00:35:07,760 --> 00:35:10,400 We're a company that's designed to create communities 563 00:35:10,400 --> 00:35:13,280 and let those communities make a difference in the world. 564 00:35:13,280 --> 00:35:16,120 We think about money as a secondary thing. 565 00:35:19,160 --> 00:35:22,400 With a revenue of over $55 billion, 566 00:35:22,400 --> 00:35:26,280 Facebook have the power to scoop up the best engineers on the planet 567 00:35:26,280 --> 00:35:28,680 with their own unique approach to recruitment. 568 00:35:31,240 --> 00:35:35,640 New employees are called "noobs" and go through a six-week boot camp 569 00:35:35,640 --> 00:35:38,240 where they're introduced to the many departments. 570 00:35:40,680 --> 00:35:42,480 This face-tracking technology 571 00:35:42,480 --> 00:35:45,360 allows us to know exactly where your eyes are, 572 00:35:45,360 --> 00:35:50,240 and that allowed you to actually superimpose funny glasses, 573 00:35:50,240 --> 00:35:52,640 or anything you really wanted. 574 00:35:52,640 --> 00:35:55,720 Most noobs don't apply for a specific job. 575 00:35:55,720 --> 00:35:58,400 They're allowed to choose where they work. 576 00:35:58,400 --> 00:36:01,400 It's supposed to make sure they end up taking on problems 577 00:36:01,400 --> 00:36:03,200 they care about most, 578 00:36:03,200 --> 00:36:08,480 but it means each department has to pitch to prospective employees. 579 00:36:08,480 --> 00:36:11,640 It's an unfortunate fact that there are a lot of people 580 00:36:11,640 --> 00:36:14,520 who are trying to reach out to minors 581 00:36:14,520 --> 00:36:17,040 on all kinds of platforms, including ours, 582 00:36:17,040 --> 00:36:19,480 for inappropriate purposes. 583 00:36:19,480 --> 00:36:21,920 We also have to worry about all kinds of things 584 00:36:21,920 --> 00:36:24,240 where people give signals on the platform 585 00:36:24,240 --> 00:36:26,960 that something terrible might happen. 586 00:36:29,160 --> 00:36:31,600 The most interesting thing here is just the mind-set 587 00:36:31,600 --> 00:36:33,240 is completely different. 588 00:36:33,240 --> 00:36:35,880 Engineers are the ones driving the change, 589 00:36:35,880 --> 00:36:39,600 whereas, in other companies, usually, it's the managers 590 00:36:39,600 --> 00:36:41,400 passing the change downwards. 591 00:36:41,400 --> 00:36:44,240 It's a big difference. I have to get used to it. 592 00:36:45,680 --> 00:36:47,120 You know, it's Facebook, 593 00:36:47,120 --> 00:36:50,120 so everything you do here is on such scale. 594 00:36:50,120 --> 00:36:54,360 Whatever small change you make is going to reach so many people. 595 00:36:54,360 --> 00:36:55,840 You need to think how you make - 596 00:36:55,840 --> 00:36:58,440 well, it's kind of a cliche - but the impact. 597 00:37:01,720 --> 00:37:04,880 Big smiles. Three, two, one. CAMERA SHUTTER CLICKS 598 00:37:04,880 --> 00:37:06,400 Awesome. 599 00:37:21,640 --> 00:37:24,080 NEWSREADER: Facebook has recorded the biggest fall 600 00:37:24,080 --> 00:37:26,800 in corporate history as shares dropped by 20%, 601 00:37:26,800 --> 00:37:30,400 wiping about $120 billion off the company's market value. 602 00:37:30,400 --> 00:37:33,240 It comes after Facebook released its latest earnings report 603 00:37:33,240 --> 00:37:35,640 and warned of decelerating revenue growth. 604 00:37:35,640 --> 00:37:38,280 User growth is slowing and the amount that it's making 605 00:37:38,280 --> 00:37:41,920 in advertising revenue per user is also declining. 606 00:37:43,920 --> 00:37:46,280 At the end of July, Facebook announced that, 607 00:37:46,280 --> 00:37:49,040 for the first time in its 14-year history, 608 00:37:49,040 --> 00:37:51,080 user growth was slowing. 609 00:37:52,680 --> 00:37:56,160 The news immediately wiped $119 billion 610 00:37:56,160 --> 00:37:59,240 off the company's market value - 611 00:37:59,240 --> 00:38:01,680 the most dramatic single-day share price decline 612 00:38:01,680 --> 00:38:03,680 in US stock market history. 613 00:38:05,920 --> 00:38:07,960 At the same time, 614 00:38:07,960 --> 00:38:11,680 Facebook are spending billions building eight new data centres 615 00:38:11,680 --> 00:38:14,560 to double capacity in the next few years. 616 00:38:17,840 --> 00:38:21,760 We're about five and a half months in on this project here. 617 00:38:21,760 --> 00:38:24,280 The exterior skins are going on. 618 00:38:26,920 --> 00:38:30,360 They're just progressing one right behind the next. 619 00:38:31,880 --> 00:38:34,960 To improve connectivity between data centres, 620 00:38:34,960 --> 00:38:39,000 they're laying thousands of miles of fibre optic cable under the sea. 621 00:38:40,520 --> 00:38:44,120 They're also constructing offices in almost every continent 622 00:38:44,120 --> 00:38:46,600 to house their rapidly growing workforce. 623 00:38:47,960 --> 00:38:50,040 We've got big projects going on 624 00:38:50,040 --> 00:38:53,840 in Hyderabad, Bangalore, Singapore, Dublin, 625 00:38:53,840 --> 00:38:55,680 New York, Austin. 626 00:38:55,680 --> 00:38:58,520 We've probably got upwards of 60 projects going on 627 00:38:58,520 --> 00:39:00,360 around the world at the moment. 628 00:39:01,760 --> 00:39:06,560 Their financial responsibilities are huge and increasing every day. 629 00:39:09,080 --> 00:39:11,360 But in spite of everything over the past year, 630 00:39:11,360 --> 00:39:13,360 they're still determined to grow... 631 00:39:16,000 --> 00:39:18,240 ..and they plan to use their connectivity 632 00:39:18,240 --> 00:39:21,040 to push into some unexpected areas. 633 00:39:36,720 --> 00:39:38,760 Facebook is built around trends 634 00:39:38,760 --> 00:39:41,560 in the way users interact on their platform. 635 00:39:43,200 --> 00:39:44,800 It's a way of building things 636 00:39:44,800 --> 00:39:47,280 they believe people really want or need... 637 00:39:48,720 --> 00:39:52,560 ..making it a more essential part of people's everyday lives. 638 00:39:56,200 --> 00:40:01,400 This entire area out here is the Social Good team. 639 00:40:01,400 --> 00:40:03,960 We have charitable giving, 640 00:40:03,960 --> 00:40:07,120 we have mentorship, we have crisis response, 641 00:40:07,120 --> 00:40:12,800 and this part out here is the blood donations team 642 00:40:14,160 --> 00:40:18,640 Hema runs a team in a department called Social Good. 643 00:40:18,640 --> 00:40:21,400 They're building products that tackle health issues 644 00:40:21,400 --> 00:40:25,360 like blood shortage - a big problem in the developing world. 645 00:40:28,160 --> 00:40:30,960 According to the World Health Organization, 646 00:40:30,960 --> 00:40:33,640 there are 70 countries with a shortage of blood... 647 00:40:35,040 --> 00:40:37,640 ..where those in need have to look for their own. 648 00:40:41,640 --> 00:40:43,920 So, Hema and her team are trying to see 649 00:40:43,920 --> 00:40:48,000 if they can find a way to solve this problem worldwide 650 00:40:48,000 --> 00:40:49,840 using Facebook. 651 00:40:53,680 --> 00:40:56,040 The highest shortage manifests itself 652 00:40:56,040 --> 00:40:58,040 on platforms like ours 653 00:40:58,040 --> 00:41:02,000 as people looking for other people to donate blood. 654 00:41:02,000 --> 00:41:03,640 It's an emergency reaction 655 00:41:03,640 --> 00:41:06,360 of people who are put under stressful conditions 656 00:41:06,360 --> 00:41:08,920 where they have to go find their own blood donors. 657 00:41:08,920 --> 00:41:12,080 I call it the BYOB system - Bring Your Own Blood. 658 00:41:12,080 --> 00:41:16,440 But that's not how countries want to operate. 659 00:41:16,440 --> 00:41:18,520 Sustainable supply comes from 660 00:41:18,520 --> 00:41:21,720 how do we connect these donors to opportunities? 661 00:41:21,720 --> 00:41:23,600 I see the problem statement as, 662 00:41:23,600 --> 00:41:26,800 country X needs 1% of its population to donate. 663 00:41:26,800 --> 00:41:31,000 That's a problem. Country X isn't achieving that outcome. 664 00:41:31,000 --> 00:41:34,120 We're going to try to help country X solve that problem. 665 00:41:36,280 --> 00:41:40,000 About five years ago, my dad was diagnosed with 666 00:41:40,000 --> 00:41:42,040 lung and liver cancer. 667 00:41:42,040 --> 00:41:45,560 He was losing about a litre of blood every day. 668 00:41:45,560 --> 00:41:48,360 And the doctors in the hospital were like, 669 00:41:48,360 --> 00:41:52,320 "We're going to do our best to help find him units of blood, 670 00:41:52,320 --> 00:41:56,440 "but we really will need more donors to come in." 671 00:41:56,440 --> 00:41:58,320 My siblings and I got on the phone 672 00:41:58,320 --> 00:42:02,360 and called every friend and every family member that we knew of, 673 00:42:02,360 --> 00:42:04,600 and they called other people. 674 00:42:04,600 --> 00:42:08,040 We eventually lost him three months later, 675 00:42:08,040 --> 00:42:11,280 but that left a pretty strong mark for me. 676 00:42:12,840 --> 00:42:18,000 We want 1% of each country's population 677 00:42:18,000 --> 00:42:22,160 to be donating every year, or however we calculate that. 678 00:42:22,160 --> 00:42:25,720 But there would be a metric and it would be based on donations. 679 00:42:25,720 --> 00:42:29,160 Let's imagine we could create a real-time efficient market for blood 680 00:42:29,160 --> 00:42:31,480 where all the demand was actually 681 00:42:31,480 --> 00:42:33,840 accurately expressed on the platform, 682 00:42:33,840 --> 00:42:36,640 and you didn't have to wait eight years for the WHO to tell you 683 00:42:36,640 --> 00:42:40,320 what the blood need was in Pakistan two years ago. 684 00:42:40,320 --> 00:42:45,000 You'd just be like, "Oh, Pakistan needed 800,000 units last year 685 00:42:45,000 --> 00:42:46,840 "and it was all on Facebook." 686 00:42:46,840 --> 00:42:49,520 I mean, you know, like, aspirational goal. 687 00:42:51,720 --> 00:42:54,760 The team are going to build a blood donation hub 688 00:42:54,760 --> 00:42:57,960 to connect potential donors to those in need, 689 00:42:57,960 --> 00:43:01,240 which will appear in people's News Feeds in India. 690 00:43:02,560 --> 00:43:04,920 They're not charging for the service, 691 00:43:04,920 --> 00:43:08,160 but you can't access it without a Facebook account. 692 00:43:11,680 --> 00:43:14,040 INTERVIEWER: When you try and make the world better, 693 00:43:14,040 --> 00:43:17,160 it comes with a risk you'll make it worse. 694 00:43:17,160 --> 00:43:20,080 How do you feel about the risks and responsibilities 695 00:43:20,080 --> 00:43:22,400 of taking on big social problems? 696 00:43:26,680 --> 00:43:29,280 The social problem, depending on what it is, 697 00:43:29,280 --> 00:43:32,000 IS the risk to the world. 698 00:43:32,000 --> 00:43:34,760 So, we've got to do a better job 699 00:43:34,760 --> 00:43:38,040 of making sure that we don't let bad actors use the platform 700 00:43:38,040 --> 00:43:40,880 in ways that'll harm the social fabric. 701 00:43:42,920 --> 00:43:46,800 One philosophy to approach this is, are we going to stop doing good 702 00:43:46,800 --> 00:43:49,720 because there's going to be a little bad in the world? 703 00:43:49,720 --> 00:43:51,640 I don't think so. I don't think we should. 704 00:43:53,680 --> 00:43:57,800 Building services that meet more fundamental human needs 705 00:43:57,800 --> 00:44:01,160 also makes Facebook more of an essential tool 706 00:44:01,160 --> 00:44:03,600 and gives users a reason to stay. 707 00:44:15,400 --> 00:44:19,640 Like a genie released from a bottle, Facebook have built a tool 708 00:44:19,640 --> 00:44:24,240 that's allowed the world to move and connect faster than ever before. 709 00:44:26,600 --> 00:44:29,520 But in doing so, old laws and systems 710 00:44:29,520 --> 00:44:33,600 built over hundreds of years are struggling to keep up... 711 00:44:36,640 --> 00:44:39,680 ..and yet Facebook is looking for even more ways 712 00:44:39,680 --> 00:44:42,720 to get the last remaining humans online. 713 00:44:47,600 --> 00:44:49,680 This is called Terragraph. 714 00:44:49,680 --> 00:44:52,920 So, this is an outdoor enclosure Wi-Fi routeing system 715 00:44:52,920 --> 00:44:54,760 using millimetre wave technology. 716 00:44:54,760 --> 00:44:57,240 We have this deployed in a few cities around the world. 717 00:44:57,240 --> 00:45:00,120 You can log in, for example, to San Jose right down the street 718 00:45:00,120 --> 00:45:03,520 and get 1.06 gigabits per second, up and down, for free. 719 00:45:05,600 --> 00:45:08,600 This military-grade wireless technology 720 00:45:08,600 --> 00:45:12,920 can transmit the world's fastest internet across entire cities. 721 00:45:14,480 --> 00:45:18,520 They've also been building satellites and high-altitude drones 722 00:45:18,520 --> 00:45:21,400 to beam the internet to remote places 723 00:45:21,400 --> 00:45:23,440 where most of the developing world live. 724 00:45:28,240 --> 00:45:31,120 If you study civilisations throughout the world, 725 00:45:31,120 --> 00:45:33,280 access to information is actually key. 726 00:45:33,280 --> 00:45:37,680 When you understand how other villages are governing themselves, 727 00:45:37,680 --> 00:45:39,320 you start to have more information 728 00:45:39,320 --> 00:45:41,560 about what's going to work for you and what's not. 729 00:45:41,560 --> 00:45:44,760 Sexual health awareness, so you can get rid of some of the disease 730 00:45:44,760 --> 00:45:47,760 and empower women around their ability to control their bodies. 731 00:45:47,760 --> 00:45:50,960 Low-irrigation farming techniques - how valuable would that be 732 00:45:50,960 --> 00:45:53,480 to villagers in some developing country? 733 00:45:53,480 --> 00:45:56,560 Now they're able to more efficiently use their resources 734 00:45:56,560 --> 00:45:59,600 to feed their villages, feed their populations. 735 00:45:59,600 --> 00:46:02,680 By providing faster internet for free, 736 00:46:02,680 --> 00:46:05,280 they're removing barriers to information, 737 00:46:05,280 --> 00:46:08,960 but that also allows more people to access their apps. 738 00:46:11,400 --> 00:46:14,440 Everywhere, teams across Facebook are pushing ahead, 739 00:46:14,440 --> 00:46:18,280 finding ever more ways to make the platform indispensable. 740 00:46:19,520 --> 00:46:22,160 All right, everybody, we've got an AMBER Alert activation. 741 00:46:22,160 --> 00:46:24,160 All right, we've got an AMBER Alert. Listen up. 742 00:46:24,160 --> 00:46:26,840 Generate the activity report for the documentation, 743 00:46:26,840 --> 00:46:29,240 send out the SMS notification to the leadership team, 744 00:46:29,240 --> 00:46:30,680 let them know. 745 00:46:30,680 --> 00:46:34,520 Facebook operates a centre for global security 746 00:46:34,520 --> 00:46:38,400 where they're able to directly send alerts into people's News Feeds. 747 00:46:41,040 --> 00:46:45,320 They regularly run drills to help search for missing children. 748 00:46:45,320 --> 00:46:49,080 It's a tool on which the police are increasingly reliant. 749 00:46:50,800 --> 00:46:52,240 At least once a day, 750 00:46:52,240 --> 00:46:54,640 sometimes several times a day, we'll have an alert. 751 00:46:54,640 --> 00:46:56,520 They usually come in from the police department 752 00:46:56,520 --> 00:46:57,920 where it's reported to. 753 00:46:57,920 --> 00:47:00,800 All right, KP's good. Get approval, please. 754 00:47:02,000 --> 00:47:03,200 You're good. 755 00:47:03,200 --> 00:47:06,440 Obviously, law enforcement has their ways of pushing stuff out to people, 756 00:47:06,440 --> 00:47:08,280 but not the kind of access we have to people. 757 00:47:08,280 --> 00:47:11,200 You know, with the billions of users we have around the world, 758 00:47:11,200 --> 00:47:13,840 it's quite a powerful tool. 759 00:47:16,120 --> 00:47:20,480 Because Facebook gather user data like location status, 760 00:47:20,480 --> 00:47:25,000 they can let people know when a child has gone missing near them. 761 00:47:25,000 --> 00:47:26,920 When you've got a child's life on the line, 762 00:47:26,920 --> 00:47:28,760 we want to be as quick as humanly possible. 763 00:47:28,760 --> 00:47:31,320 All right, I'll push this out to public. 764 00:47:31,320 --> 00:47:33,000 HE EXHALES 765 00:47:34,680 --> 00:47:36,920 All clear. Good job, team. 766 00:47:38,280 --> 00:47:40,120 If you're a general public user right now 767 00:47:40,120 --> 00:47:42,000 and you're in this region where we sent this to - 768 00:47:42,000 --> 00:47:43,840 obviously, it's a drill, it's not going out - 769 00:47:43,840 --> 00:47:45,840 but if it did, everybody would now get the alert 770 00:47:45,840 --> 00:47:47,880 and they'd see the same thing I just saw. 771 00:47:47,880 --> 00:47:51,360 Hopefully there's enough information there for someone to find them. 772 00:47:54,200 --> 00:47:58,280 It's not known how many children have been found using this system, 773 00:47:58,280 --> 00:48:03,320 however, police call Facebook for an alert almost every day. 774 00:48:03,320 --> 00:48:06,480 By becoming a kind of emergency service, 775 00:48:06,480 --> 00:48:09,120 they're providing an essential tool. 776 00:48:11,040 --> 00:48:14,280 Facebook offers access to billions of people, 777 00:48:14,280 --> 00:48:17,560 but these services are in the hands of a private company. 778 00:48:18,920 --> 00:48:21,520 It raises questions about what activity, 779 00:48:21,520 --> 00:48:23,800 if any, should be off limits. 780 00:48:38,680 --> 00:48:41,920 After months of planning, Facebook are about to launch 781 00:48:41,920 --> 00:48:44,760 its new blood donations hub in India. 782 00:48:47,800 --> 00:48:49,520 With the click of a button, 783 00:48:49,520 --> 00:48:53,280 300 million people will be able to connect to donation centres 784 00:48:53,280 --> 00:48:57,360 that have a shortage of blood all over the country, 785 00:48:57,360 --> 00:49:00,960 and maybe it will encourage some of the other billion people in India 786 00:49:00,960 --> 00:49:02,400 to sign up, too. 787 00:49:06,120 --> 00:49:08,960 OK, this is what the blood donation hub looks like. 788 00:49:08,960 --> 00:49:11,960 At the top is a map of the user's city 789 00:49:11,960 --> 00:49:14,600 and dots indicating all the opportunities 790 00:49:14,600 --> 00:49:17,040 for them to donate nearby, but... 791 00:49:18,680 --> 00:49:22,600 ..we also have requests which are from individuals. 792 00:49:22,600 --> 00:49:25,360 Either they're in need of blood, 793 00:49:25,360 --> 00:49:28,080 or often it's their family member in need of blood, 794 00:49:28,080 --> 00:49:30,880 and they'll post using our tools 795 00:49:30,880 --> 00:49:33,320 and we can notify donors in our database 796 00:49:33,320 --> 00:49:35,240 who have the same blood type 797 00:49:35,240 --> 00:49:38,640 and are within a certain geographic area, 798 00:49:38,640 --> 00:49:41,120 and who have indicated to us that they want to donate 799 00:49:41,120 --> 00:49:42,440 and they want to help. 800 00:49:43,680 --> 00:49:45,680 In the US, we take it for granted that, 801 00:49:45,680 --> 00:49:47,920 if you go to the hospital and you need surgery, 802 00:49:47,920 --> 00:49:49,560 the blood will be there for you. 803 00:49:56,440 --> 00:49:59,640 As India begins to wake up, they launch their new hub... 804 00:50:01,320 --> 00:50:04,360 ..and, within minutes, users begin to respond. 805 00:50:08,240 --> 00:50:12,520 Come on, guys. Share, share, share. Whoa! Whoa! There she is. 806 00:50:12,520 --> 00:50:15,120 Oh, man! Where are we at? 807 00:50:15,120 --> 00:50:19,880 So, there's been 15,000 hub views so far. 15,000 hub views? 808 00:50:19,880 --> 00:50:21,760 Oh, man! 809 00:50:21,760 --> 00:50:24,920 Not bad. Not bad at all. 810 00:50:24,920 --> 00:50:27,320 People are sharing it, they're commenting on it, 811 00:50:27,320 --> 00:50:30,640 they're liking it. Like, it's really blowing up. 812 00:50:30,640 --> 00:50:34,800 Hey, it was worth building it. It was definitely worth building it. 813 00:50:34,800 --> 00:50:37,080 Since launching the new feature, 814 00:50:37,080 --> 00:50:39,720 more than 11 million people have signed up 815 00:50:39,720 --> 00:50:42,680 to donate blood on Facebook. 816 00:50:42,680 --> 00:50:45,200 I believe there should never be anybody 817 00:50:45,200 --> 00:50:47,640 who would have to go look for a blood donor ever 818 00:50:47,640 --> 00:50:49,880 because there always exists safe blood 819 00:50:49,880 --> 00:50:51,920 in their blood banks near them. 820 00:50:53,960 --> 00:50:56,760 Something that would have changed my life five years ago. 821 00:51:00,640 --> 00:51:05,120 Initiatives like these have the potential to save lives, 822 00:51:05,120 --> 00:51:09,240 but already there have been reports that the service could be exploited 823 00:51:09,240 --> 00:51:11,000 by the black market. 824 00:51:17,280 --> 00:51:21,360 Several public health officials in India warn that the tech project, 825 00:51:21,360 --> 00:51:23,000 although well-meaning, 826 00:51:23,000 --> 00:51:26,640 risks fuelling a dangerous black market for blood 827 00:51:26,640 --> 00:51:29,680 and harming the country's fragile blood collection system. 828 00:51:36,680 --> 00:51:41,480 Facebook has a huge capacity to encourage others to do good, 829 00:51:41,480 --> 00:51:45,520 but its recent history of scandals raises further questions 830 00:51:45,520 --> 00:51:49,600 as to how safe we are sharing sensitive data with them. 831 00:51:50,840 --> 00:51:53,480 And as they continue to push forwards, 832 00:51:53,480 --> 00:51:56,360 the scandals show no sign of slowing down. 833 00:51:57,600 --> 00:52:01,120 The Business Secretary Greg Clark has called a security breach 834 00:52:01,120 --> 00:52:04,840 affecting millions of Facebook users "a very worrying development". 835 00:52:04,840 --> 00:52:08,480 At least 50 million Facebook users have had their accounts hacked. 836 00:52:08,480 --> 00:52:12,960 Another huge security breach at the world's biggest social network. 837 00:52:15,880 --> 00:52:19,040 The latest security hack raises huge questions 838 00:52:19,040 --> 00:52:22,240 about the company's ability to protect the user data 839 00:52:22,240 --> 00:52:24,920 of nearly a third of the world's population... 840 00:52:26,200 --> 00:52:27,840 ..and it throws into question 841 00:52:27,840 --> 00:52:31,560 the effectiveness of their efforts to fix Facebook so far. 842 00:52:45,040 --> 00:52:49,280 Every single day, they're facing more and more of these challenges. 843 00:52:51,600 --> 00:52:54,400 Overnight, some behaviour they'd never seen before 844 00:52:54,400 --> 00:52:56,120 was detected on the network, 845 00:52:56,120 --> 00:52:58,600 so Vlad has called an urgent meeting. 846 00:53:00,080 --> 00:53:02,920 Users around the world were getting friend requests 847 00:53:02,920 --> 00:53:05,760 from someone they didn't know - a fake account. 848 00:53:06,920 --> 00:53:10,120 But this wasn't stopped by their artificial intelligence 849 00:53:10,120 --> 00:53:13,360 because the account behaved in an unusual way. 850 00:53:15,240 --> 00:53:17,240 So, what happened was 851 00:53:17,240 --> 00:53:21,160 the profile changed its name, started friending a bunch of people, 852 00:53:21,160 --> 00:53:22,840 and you can see friend rejects... 853 00:53:22,840 --> 00:53:24,920 A lot of them... I mean, a lot of these were rejects. 854 00:53:24,920 --> 00:53:28,120 And immediately after that, it changed it back. 855 00:53:28,120 --> 00:53:31,560 It appears the fake account changed its profile name 856 00:53:31,560 --> 00:53:35,240 and gender twice to try to avoid detection. 857 00:53:35,240 --> 00:53:38,480 So, they're switching and then switching back? Yes. 858 00:53:38,480 --> 00:53:40,480 This is kind of interesting in a sense that 859 00:53:40,480 --> 00:53:42,360 they're not throwing away their account. 860 00:53:42,360 --> 00:53:44,760 Usually, it used to be they would go in, 861 00:53:44,760 --> 00:53:48,640 you know, create an account, make it sort of look reasonable, 862 00:53:48,640 --> 00:53:50,840 and then start scamming, 863 00:53:50,840 --> 00:53:53,080 and then, you know, we would take it down. 864 00:53:53,080 --> 00:53:56,760 Now they're taking it back to make it look reasonable again, 865 00:53:56,760 --> 00:53:58,160 which is weird. 866 00:53:59,400 --> 00:54:02,040 Vlad's worried this could be organised criminals 867 00:54:02,040 --> 00:54:05,920 trying to see if they can find a weakness in the system. 868 00:54:05,920 --> 00:54:08,280 If they can avoid detection with one account, 869 00:54:08,280 --> 00:54:12,520 they could do it en masse and start posting scams on the platform. 870 00:54:13,840 --> 00:54:16,280 First of all, we need to figure out if this is common enough, 871 00:54:16,280 --> 00:54:19,920 and then we should figure out what's the underlying pattern behind this. 872 00:54:19,920 --> 00:54:22,360 Sometimes, a back door gets discovered, 873 00:54:22,360 --> 00:54:25,800 and then it blows up, the same like they discovered BLEEP, 874 00:54:25,800 --> 00:54:28,240 and then just follow the magic recipe. 875 00:54:28,240 --> 00:54:31,920 If you assume that there's a whole organisation behind this 876 00:54:31,920 --> 00:54:34,920 because they have lots of people you need to organise and all that, 877 00:54:34,920 --> 00:54:38,920 then you end up kind of finding those structures, 878 00:54:38,920 --> 00:54:41,440 and that's what you really want to go after. 879 00:54:42,640 --> 00:54:44,680 Cool. Thanks, everyone. Thank you. 880 00:54:49,040 --> 00:54:52,640 By studying this fake account and its behaviour, 881 00:54:52,640 --> 00:54:55,960 they should be able to retrain their artificial intelligence 882 00:54:55,960 --> 00:54:59,720 to block any others like them and, in doing so, 883 00:54:59,720 --> 00:55:03,120 hopefully prevent any users from being scammed. 884 00:55:05,600 --> 00:55:08,000 INTERVIEWER: Do you take this stuff home 885 00:55:08,000 --> 00:55:10,280 or are you able to switch off? 886 00:55:10,280 --> 00:55:12,040 It's hard sometimes. 887 00:55:12,040 --> 00:55:16,360 You know, sometimes, you go home, and, you know, some days, 888 00:55:16,360 --> 00:55:21,200 like, it's a rough day and you'll feel it at home. 889 00:55:33,240 --> 00:55:37,680 2018 was the year that rocked Facebook to its core 890 00:55:37,680 --> 00:55:41,560 and woke the world up to its reach and influence over society. 891 00:55:42,880 --> 00:55:46,280 A year on, they've announced plans to fundamentally shift 892 00:55:46,280 --> 00:55:50,920 how Facebook works by making it more focused on user privacy. 893 00:55:51,960 --> 00:55:54,960 Facebook has a history of promising things 894 00:55:54,960 --> 00:55:58,360 when regulators come knocking, when critics start writing, 895 00:55:58,360 --> 00:56:00,640 when reporters come to their door, 896 00:56:00,640 --> 00:56:03,600 promising things in terms of safety and in terms of privacy 897 00:56:03,600 --> 00:56:05,320 and then not delivering on them. 898 00:56:06,680 --> 00:56:09,280 Whether or not they deliver on their promises, 899 00:56:09,280 --> 00:56:13,760 it's clear that whoever has control over our personal data 900 00:56:13,760 --> 00:56:16,920 also holds the key to immense power. 901 00:56:18,920 --> 00:56:21,000 This raises a bigger question 902 00:56:21,000 --> 00:56:24,160 of who decides what is right and wrong 903 00:56:24,160 --> 00:56:26,240 in a fast-evolving world. 904 00:56:27,360 --> 00:56:29,520 Whether it's because of an event in the world 905 00:56:29,520 --> 00:56:31,520 or an election that's happening 906 00:56:31,520 --> 00:56:35,200 or some sort of new trend among young people, 907 00:56:35,200 --> 00:56:37,840 there's always something that makes us think, 908 00:56:37,840 --> 00:56:40,080 "Do we have these lines in the right place?" 909 00:56:41,280 --> 00:56:45,680 What kinds of information are users comfortable sharing? 910 00:56:45,680 --> 00:56:48,360 And how much do they deserve to know about the ways 911 00:56:48,360 --> 00:56:51,680 in which their data is used? 912 00:56:51,680 --> 00:56:55,520 Whether it will ultimately become a force for good or bad 913 00:56:55,520 --> 00:56:59,200 depends now on Facebook's ability to regain control 914 00:56:59,200 --> 00:57:01,760 of the genie they've released on the world. 915 00:57:03,480 --> 00:57:05,280 I really liken it to when 916 00:57:05,280 --> 00:57:08,720 we were building railroads for the first time. 917 00:57:08,720 --> 00:57:11,960 We built the railroads because we wanted to get from A to B. 918 00:57:11,960 --> 00:57:15,880 We actually had no idea the economies, 919 00:57:15,880 --> 00:57:18,640 the societal growth that would happen 920 00:57:18,640 --> 00:57:21,480 by building those connections. 921 00:57:21,480 --> 00:57:24,560 We're at a very similar place with the internet, 922 00:57:24,560 --> 00:57:29,120 with being able to share information around the world in moments. 923 00:57:29,120 --> 00:57:31,440 How do you maximise all of those good things 924 00:57:31,440 --> 00:57:33,080 that come from connecting people 925 00:57:33,080 --> 00:57:35,360 while also mitigating some of the negativity 926 00:57:35,360 --> 00:57:36,960 that can come alongside it? 927 00:57:36,960 --> 00:57:39,200 It will take a collective effort of many people 928 00:57:39,200 --> 00:57:41,400 across private sector and public sector 929 00:57:41,400 --> 00:57:42,840 and academia and government 930 00:57:42,840 --> 00:57:45,440 to come together and raise the important questions. 931 00:57:45,440 --> 00:57:49,600 How should we operate a society when everyone really does have a voice? 932 00:57:49,600 --> 00:57:54,880 The last year has been challenging, but it's also been the year where 933 00:57:54,880 --> 00:57:58,880 we went from being this naively optimistic child 934 00:57:58,880 --> 00:58:00,920 to being a slightly more mature adult. 935 00:58:02,640 --> 00:58:05,200 And we're now on this path of, 936 00:58:05,200 --> 00:58:07,680 you know, accepting our responsibility. 937 00:58:09,040 --> 00:58:11,480 You know, this is a very new world. 938 00:58:11,480 --> 00:58:13,800 Everything we're doing is brand-new... 939 00:58:15,160 --> 00:58:17,760 ..and that comes with painful learnings.