1 00:00:00,000 --> 00:00:14,990 *34c3 intro* 2 00:00:14,990 --> 00:00:23,740 Herald: the next talk is Marloes de Valk, she's an artist and writer from the 3 00:00:23,740 --> 00:00:31,289 Netherlands and she's working with lots of different materials and media and at the 4 00:00:31,289 --> 00:00:40,490 moment she's doing an 8-bit game, so the topic is "why do we anthropomorphize 5 00:00:40,490 --> 00:00:48,460 computers and dehumanize ourselves in the process?" and we have a mumble, which is 6 00:00:48,460 --> 00:00:54,240 doing the translation, the talk is in English and we will translate into French 7 00:00:54,240 --> 00:01:01,930 and German. Okay, give a big applause for Marloes! 8 00:01:01,930 --> 00:01:09,670 *applause* 9 00:01:09,670 --> 00:01:13,740 Marloes: Thank you and thank you all for 10 00:01:13,740 --> 00:01:20,860 coming, my name is Marloes de Valk and I'm going to talk about anthropomorphization 11 00:01:20,860 --> 00:01:27,450 and I will approach this as a survival strategy, see how it works and if it is 12 00:01:27,450 --> 00:01:34,490 effective. And when I'm speaking of big data, which is an umbrella term, my focus 13 00:01:34,490 --> 00:01:38,940 will be on the socio-technical aspect of the phenomenon, the assumptions and 14 00:01:38,940 --> 00:01:44,090 beliefs surrounding Big Data and on research using data exhaust or found data 15 00:01:44,090 --> 00:01:49,770 such as status updates on social media web searches and credit card payments. 16 00:01:55,750 --> 00:02:08,300 Oh and now my slides are frozen. Oh my gosh. 17 00:02:09,460 --> 00:02:11,280 Audience: Have you tried turning it of and on again? 18 00:02:11,280 --> 00:02:15,880 Marloes: *laughs* I will in a moment. gosh it's 19 00:02:15,880 --> 00:02:25,730 completely frozen... I'm very sorry, technical staff I have to exit, if I can. 20 00:02:25,730 --> 00:02:40,730 I can't. Help! I have to get rid of something I think, should we just kill it? 21 00:02:40,730 --> 00:02:55,280 That's so stupid yeah. 22 00:02:55,280 --> 00:02:58,280 But they're gonna have a coffee soon and then it's gonna 23 00:02:58,280 --> 00:03:33,070 Yes, force quit... I think I know what the problem is. I'm sorry it's, it's 24 00:03:33,070 --> 00:04:09,940 really not working. All right let's see if we're back. 25 00:04:09,940 --> 00:04:18,870 Okay, okay so sorry for the interruption. 26 00:04:18,870 --> 00:04:24,150 I wanted to start by letting Silicon Valley itself tell a story about 27 00:04:24,150 --> 00:04:30,930 technology, really, sorry about the interruption. So, Silicon Valley propaganda 28 00:04:30,930 --> 00:04:35,330 during our lifetime, we're about to see the transformation of the human race, it's 29 00:04:35,330 --> 00:04:39,400 really something that blows my mind every time I think about it. People have no idea 30 00:04:39,400 --> 00:04:42,800 how fast the world is changing and I want to give you a sense of that because it 31 00:04:42,800 --> 00:04:47,650 fills me with awe and with an extraordinary sense of responsibility. I 32 00:04:47,650 --> 00:04:52,120 want to give you a sense of why now is different why this decade the next decade 33 00:04:52,120 --> 00:04:56,370 is not interesting times but THE most extraordinary times ever in 34 00:04:56,370 --> 00:05:00,520 human history and they truly are. What we're talking about here is the notion 35 00:05:00,520 --> 00:05:04,670 that faster cheaper computing power which is almost like a force of nature, is 36 00:05:04,670 --> 00:05:09,350 driving a whole slew of technologies, technology being the force that takes what 37 00:05:09,350 --> 00:05:14,090 used to be scarce and make it abundant. That is why we're heading towards this 38 00:05:14,090 --> 00:05:19,630 extraordinary age of abundance. The future will not take care of itself as we know 39 00:05:19,630 --> 00:05:23,590 the world looks to America for progress and America looks to California and if you 40 00:05:23,590 --> 00:05:28,070 ask most Californians where they get their progress they'll point towards the bay, 41 00:05:28,070 --> 00:05:33,210 but here at the bay there is no place left to point, so we have to create solutions 42 00:05:33,210 --> 00:05:37,840 and my goal is to simplify complexity, take Internet technology and cross it with 43 00:05:37,840 --> 00:05:43,020 an old industry and magic and progress and big things can happen. I really think 44 00:05:43,020 --> 00:05:46,990 there are two fundamental paths for humans, one path is we stay on earth 45 00:05:46,990 --> 00:05:51,650 forever, or some eventual extinction event wipes us out, I don't have a doomsday 46 00:05:51,650 --> 00:05:56,009 prophecy but history suggests some doomsday event will happen. The 47 00:05:56,009 --> 00:06:00,930 alternative is becoming a spacefaring and multiplanetary species and it will be like 48 00:06:00,930 --> 00:06:06,670 really fun to go, you'll have a great time. We will set on Mars and we should, 49 00:06:06,670 --> 00:06:10,630 because it's cool. When it comes to space I see it as my job to build infrastructure 50 00:06:10,630 --> 00:06:14,480 the hard way. I'm using my resources to put in that infrastructure so that the 51 00:06:14,480 --> 00:06:19,150 next generation of people can have a dynamic entrepreneurial solar system as 52 00:06:19,150 --> 00:06:23,000 interesting as we see on the internet today. We want the population to keep 53 00:06:23,000 --> 00:06:29,430 growing on this planet, we want to keep you using more energy per capita. Death 54 00:06:29,430 --> 00:06:33,889 makes me very angry, probably the most extreme form of inequality is between 55 00:06:33,889 --> 00:06:38,180 people who are alive and people who are dead. I have the idea that aging is 56 00:06:38,180 --> 00:06:42,050 plastic, that it's encoded and if something is encoded you can crack the 57 00:06:42,050 --> 00:06:46,390 code if you can crack the code you can hack the code and thermodynamically there 58 00:06:46,390 --> 00:06:51,470 should be no reason we can't defer entropy indefinitely. We can end aging forever. 59 00:06:51,470 --> 00:06:54,510 This is not about Silicon Valley billionaires 60 00:06:54,510 --> 00:06:57,199 living forever off the blood of young people. 61 00:06:57,199 --> 00:07:02,020 It's about a Star Trek future where no one dies of preventable diseases where life is 62 00:07:02,020 --> 00:07:05,770 fair. Health technology is becoming an information technology, where we can read 63 00:07:05,770 --> 00:07:10,539 and edit our own genomes clearly it is possible through technology to make death 64 00:07:10,539 --> 00:07:16,880 optional. Yes, our bodies are information processing systems. We can enable human 65 00:07:16,880 --> 00:07:22,350 transformations that would rival Marvel Comics super muscularity ultra endurance, 66 00:07:22,350 --> 00:07:26,880 super radiation resistance, you could have people living on the moons of Jupiter, 67 00:07:26,880 --> 00:07:30,259 who'd be modified in this way and they could physically harvest energy from the 68 00:07:30,259 --> 00:07:35,199 gamma rays they were exposed to. Form a culture connected with the ideology of the 69 00:07:35,199 --> 00:07:39,080 future, promoting technical progress artificial intellects, multi-body 70 00:07:39,080 --> 00:07:44,419 immortality and cyborgization. We are at the beginning of the beginning the first 71 00:07:44,419 --> 00:07:49,250 hour of day one, there have never been more opportunities the greatest products 72 00:07:49,250 --> 00:07:54,830 of the next 25 years have not been invented yet. You are not too late. 73 00:07:54,830 --> 00:07:59,700 We're going to take over the world, one robot at a time. It's gonna be an AI that 74 00:07:59,700 --> 00:08:03,740 is able to source create solve an answer just what is your desire. I mean this is 75 00:08:03,740 --> 00:08:09,380 an almost godlike view of the future. AI is gonna be magic. Especially in the 76 00:08:09,380 --> 00:08:14,310 digital manufacturing world, what is going to be created will effectively be a god, 77 00:08:14,310 --> 00:08:17,810 the idea needs to spread before the technology, the church is how we spread 78 00:08:17,810 --> 00:08:21,690 the word, the gospel. If you believe in it, start a conversation with someone else 79 00:08:21,690 --> 00:08:26,280 and help them understand the same things. Computers are going to take over from 80 00:08:26,280 --> 00:08:29,940 humans, no question, but when I got that thinking in my head about if I'm going to 81 00:08:29,940 --> 00:08:33,599 be treated in the future as a pet to these smart machines, well I'm gonna treat my 82 00:08:33,599 --> 00:08:38,068 own pet dog really nice, but in the end we may just have created the species that is 83 00:08:38,068 --> 00:08:42,688 above us. Chaining it isn't gonna be the solution as it will be stronger than any 84 00:08:42,688 --> 00:08:48,149 change could put on. the existential risk that is associated with AI we will not be 85 00:08:48,149 --> 00:08:51,180 able to beat AI, so then as the saying goes if you 86 00:08:51,180 --> 00:08:54,879 can't beat them, join them. History has shown us we aren't gonna win 87 00:08:54,879 --> 00:08:58,879 this war by changing human behavior but maybe we can build systems that are so 88 00:08:58,879 --> 00:09:03,149 locked down, that humans lose the ability to make dumb mistakes until we gain the 89 00:09:03,149 --> 00:09:08,190 ability to upgrade the human brain, it's the only way. Let's stop pretending we can 90 00:09:08,190 --> 00:09:11,929 hold back the development of intelligence when there are clear massive short-term 91 00:09:11,929 --> 00:09:16,069 economic benefits to those who develop it and instead understand the future and have 92 00:09:16,069 --> 00:09:20,611 it treat us like a beloved elder who created it. As a company, one of our 93 00:09:20,611 --> 00:09:24,119 greatest cultural strengths is accepting the fact that if you're gonna invent, 94 00:09:24,119 --> 00:09:29,019 you're gonna disrupt. Progress is happening because there is economic 95 00:09:29,019 --> 00:09:32,629 advantage to having machines work for you and solve problems for you. People are 96 00:09:32,629 --> 00:09:38,929 chasing that. AI, the term has become more of a broad, almost marketing driven term 97 00:09:38,929 --> 00:09:42,439 and I'm probably okay with that. What matters is what people think of when they 98 00:09:42,439 --> 00:09:47,339 hear of this. We are in a deadly race between politics and technology, the fate 99 00:09:47,339 --> 00:09:51,420 of our world may depend on the effort of a single person who builds or propagates the 100 00:09:51,420 --> 00:09:57,749 machinery of freedom, that makes the world safe for capitalism. 101 00:09:57,749 --> 00:10:04,470 These were all quotes. Every single one. not only Silicon Valley CEO speak of 102 00:10:04,470 --> 00:10:07,889 Technology in mysterious ways, let's see some examples from the media. 103 00:10:07,889 --> 00:10:12,079 Our official intelligence regulation, "lets not regulate mathematics" a headline 104 00:10:12,079 --> 00:10:16,779 from import dot IO from May 2016 about the European general data protection 105 00:10:16,779 --> 00:10:22,319 regulation and the article concludes autonomous cars should be regulated as 106 00:10:22,319 --> 00:10:26,129 cars, they should safely deliver users to their destinations in the real world and 107 00:10:26,129 --> 00:10:31,320 overall reduce the number of accidents. How they achieve this is irrelevant. With 108 00:10:31,320 --> 00:10:34,299 enough data the numbers speak for themselves which comes from the super 109 00:10:34,299 --> 00:10:40,240 famous article "The end of theory" from Chris Anderson in Wired magazine 2008. 110 00:10:40,240 --> 00:10:45,309 "Google creates an AI that can teach itself to be better than humans" headline 111 00:10:45,309 --> 00:10:48,549 from "The Independent. The article continues the company's AI 112 00:10:48,549 --> 00:10:52,639 division deepmind has unveiled "alpha go zero" an extremely advanced system that 113 00:10:52,639 --> 00:10:58,160 managed to accumulate thousands of years of human knowledge within days. Microsoft 114 00:10:58,160 --> 00:11:02,850 apologizing for their teen chat Bot gone Nazi stating it wasn't their fault. "We're 115 00:11:02,850 --> 00:11:06,199 deeply sorry for the unintended and hurtful tweets from Tay which do not 116 00:11:06,199 --> 00:11:13,699 represent who we are or what we stand for nor how we design Tay" and then the PC 117 00:11:13,699 --> 00:11:18,829 world article "AI just 3d printed a brand new Rembrandt and it's shockingly good", 118 00:11:18,829 --> 00:11:21,550 the subtitle reads "the next Rembrandt project used data and 119 00:11:21,550 --> 00:11:26,860 deep learning to produce uncanny results". Advertising firm J walter thompson 120 00:11:26,860 --> 00:11:32,049 unveiled a 3d printed painting called "the next Rembrandt" based on 346 paintings of 121 00:11:32,049 --> 00:11:36,100 the old master, not just PC world, but many more articles touted similar titles 122 00:11:36,100 --> 00:11:41,209 presenting the painting to the public, as if it were made by a computer, a 3d 123 00:11:41,209 --> 00:11:45,179 printer, AI and deep learning. It is clear though, that the computer programmers who 124 00:11:45,179 --> 00:11:48,765 worked on the project are not computers and neither are the people who tagged the 125 00:11:48,765 --> 00:11:53,509 346 Rembrandt paintings by hand. The painting was made by a team of programmers 126 00:11:53,509 --> 00:11:58,589 and researchers and it took them 18 months to do. So what is communicated through 127 00:11:58,589 --> 00:12:03,130 these messages is that the computer did it, yet there is no strong AI, as in 128 00:12:03,130 --> 00:12:07,029 consciousness in machines at this moment, only very clever automation, meaning it 129 00:12:07,029 --> 00:12:11,570 was really us. We comprehend the role and function of non-human actors rationally, 130 00:12:11,570 --> 00:12:15,680 but still intuitively approach them differently. We anthropomorphize and 131 00:12:15,680 --> 00:12:19,160 stories about the intelligent things machines can do and force the belief in 132 00:12:19,160 --> 00:12:26,620 the human-like agency of machines, so why do we do it. 133 00:12:26,620 --> 00:12:28,019 I'd like to think of this as two survival 134 00:12:28,019 --> 00:12:32,680 strategies that found each other in big data and AI discourse. George Zarkadakis 135 00:12:32,680 --> 00:12:37,209 in the book "in our own image" describes the root of anthropomorphization, during 136 00:12:37,209 --> 00:12:40,610 the evolution of the modern mind humans acquired and developed general-purpose 137 00:12:40,610 --> 00:12:44,259 language, through social language and this first social language was a way of 138 00:12:44,259 --> 00:12:48,999 grooming of creating social cohesion. We gained theory of mind to believe that 139 00:12:48,999 --> 00:12:52,959 other people have thoughts, desires, intentions and feelings of their own - 140 00:12:52,959 --> 00:12:56,579 Empathy. And this led to the describing of the world in social terms, perceiving 141 00:12:56,579 --> 00:13:01,259 everything around us as agents possessing mind, including the nonhuman, when hunting 142 00:13:01,259 --> 00:13:05,600 anthropomorphizing animals had a great advantage because you could strategize, 143 00:13:05,600 --> 00:13:12,190 predict their movements. They show through multiple experiment- Oh, Reeves and Nass 144 00:13:12,190 --> 00:13:16,040 were picking up on this anthropomorphization and they show through 145 00:13:16,040 --> 00:13:20,220 multiple experiments that we haven't changed that much, through multiple 146 00:13:20,220 --> 00:13:24,309 experiments they show how people treat computers, television and new media like 147 00:13:24,309 --> 00:13:28,249 real people in places even though it test subjects were completely unaware of it, 148 00:13:28,249 --> 00:13:33,189 they responded to computers as they would to people being polite cooperative, 149 00:13:33,189 --> 00:13:37,009 attributing personality characteristics such as aggressiveness, humor, expertise 150 00:13:37,009 --> 00:13:41,339 and even gender. Meaning we haven't evolved that much, we still do it. 151 00:13:41,339 --> 00:13:44,799 Microsoft unfortunately misinterpreted their research and developed the innocent 152 00:13:44,799 --> 00:13:49,589 yet much hated Clippy the paper clip, appearing one year later in office 97. 153 00:13:49,589 --> 00:13:54,920 This survival strategy found its way into another one. The Oracle. Survival through 154 00:13:54,920 --> 00:13:58,480 predicting events. The second strategy is trying to predict 155 00:13:58,480 --> 00:14:03,049 the future, to steer events in our favor, in order to avoid disaster. The fear of 156 00:14:03,049 --> 00:14:06,989 death has inspired us throughout the ages to try and predict the future and it has 157 00:14:06,989 --> 00:14:10,480 led us to consult Oracles and to creating a new one. 158 00:14:10,480 --> 00:14:13,679 Because we cannot predict the future in the midst of lives many insecurities, we 159 00:14:13,679 --> 00:14:18,000 desperately crave the feeling of being in control over our destiny, we have 160 00:14:18,000 --> 00:14:23,269 developed ways to calm our anxiety, to comfort ourselves and what we do is we 161 00:14:23,269 --> 00:14:27,420 obfuscate that human hand in a generation of messages that require an objective or 162 00:14:27,420 --> 00:14:31,660 authority feel, although disputed is commonly believed that the Delphic Oracle 163 00:14:31,660 --> 00:14:36,079 delivered messages from her god Apollo in a state of trance induced by intoxicating 164 00:14:36,079 --> 00:14:40,949 vapors arising from the chasm over which she was seated, possesed by her God the 165 00:14:40,949 --> 00:14:45,209 Oracle spoke ecstatically and spontaneously. Priests of the temple then 166 00:14:45,209 --> 00:14:50,360 translated her gibberish into the prophesies, the seekers of advice were 167 00:14:50,360 --> 00:14:56,220 sent home with. And Apollo had spoken. Nowadays we turn to data for advice. The 168 00:14:56,220 --> 00:15:00,121 Oracle of Big Data functions in a similar way to the Oracle of Delphi. Algorithms 169 00:15:00,121 --> 00:15:04,750 programmed by humans are fed data and consequently spit out numbers that are 170 00:15:04,750 --> 00:15:07,949 then translated and interpreted by researchers into the prophecies the 171 00:15:07,949 --> 00:15:12,230 seekers of advice are sent home with. The bigger data the set, the more accurate the 172 00:15:12,230 --> 00:15:16,290 results. Data has spoken. We are brought closer to the truth, to 173 00:15:16,290 --> 00:15:22,060 reality as it is, unmediated by us, subjective biased and error-prone humans. 174 00:15:22,060 --> 00:15:26,529 This Oracle inspires great hope. It's a utopia and this is best putting words in 175 00:15:26,529 --> 00:15:29,600 the article "The end of theory" by Anderson where he states that with enough 176 00:15:29,600 --> 00:15:34,859 data the numbers can speak for themselves. We can forget about taxonomy, ontology, 177 00:15:34,859 --> 00:15:39,459 psychology, who knows why people do what they do. The point is they do it and we 178 00:15:39,459 --> 00:15:43,229 can track and measure it with unprecedented fidelity, with enough data 179 00:15:43,229 --> 00:15:48,149 the numbers speak for themselves. This Oracle is of course embraced with great 180 00:15:48,149 --> 00:15:52,649 enthusiasm by database and storage businesses as shown here in an Oracle 181 00:15:52,649 --> 00:15:58,660 presentation slide. High Five! And getting it right one out of ten times and using 182 00:15:58,660 --> 00:16:02,730 the one success story to strengthen the belief in big data superpowers happens a 183 00:16:02,730 --> 00:16:06,889 lot in the media, a peculiar example is the story on Motherboard about how 184 00:16:06,889 --> 00:16:10,579 "Cambridge Analytica" helped Trump win the elections by psychologically profiling the 185 00:16:10,579 --> 00:16:14,809 entire American population and using targeted Facebook ads to influence the 186 00:16:14,809 --> 00:16:20,209 results of the election. This story evokes the idea that they know more about you 187 00:16:20,209 --> 00:16:26,790 than your own mother. The article reads "more likes could even surpass what a 188 00:16:26,790 --> 00:16:31,019 person thought they knew about themselves" and although this form of manipulation is 189 00:16:31,019 --> 00:16:36,160 seriously scary in very undemocratic as Cathy O'Neil author of "weapons 190 00:16:36,160 --> 00:16:41,879 of mass mass destruction" notes, "don't believe the hype". 191 00:16:41,879 --> 00:16:45,459 It wasn't just Trump everyone was doing it Hillary was using the groundwork, a 192 00:16:45,459 --> 00:16:49,380 startup funded by Google's Eric Schmidt, Obama used groundwork too, but the 193 00:16:49,380 --> 00:16:53,089 groundwork somehow comes across a lot more cute compared to Cambridge analytica, 194 00:16:53,089 --> 00:16:56,279 funded by billionaire Robert Mercer who also is heavily invested in all-tried 195 00:16:56,279 --> 00:17:00,759 media outlet Breitbart, who describes itself as a killing machine waging the war 196 00:17:00,759 --> 00:17:05,869 for the West, he also donated Cambridge analytica service to the brexit campaign. 197 00:17:05,869 --> 00:17:08,949 The Motherboard article and many others describing the incredibly detailed 198 00:17:08,949 --> 00:17:13,019 knowledge Cambridge Analytica has on American citizens were amazing advertising 199 00:17:13,019 --> 00:17:17,659 for the company, but most of all a warning sign that applying big data research to 200 00:17:17,659 --> 00:17:21,539 elections creates a very undemocratic Asymmetry and available information and 201 00:17:21,539 --> 00:17:27,559 undermines the notion of an informed citizenry. Dana Boyd and Kate Crawford 202 00:17:27,559 --> 00:17:31,309 described the beliefs attached to big data as a mythology "the widespread believe 203 00:17:31,309 --> 00:17:34,880 that large datasets offer a higher form of intelligence and knowledge that can 204 00:17:34,880 --> 00:17:39,120 generate insights, that were previously impossible with the aura of truth 205 00:17:39,120 --> 00:17:43,610 objectivity and accuracy". The deconstruction of this myth was 206 00:17:43,610 --> 00:17:48,080 attempted as early as 1984 in a spreadsheet way of knowledge, Steven Levi 207 00:17:48,080 --> 00:17:51,929 describes how the authority of look of a spreadsheet and the fact that it was done 208 00:17:51,929 --> 00:17:55,330 by a computer has a strong persuasive effect on people, leading to the 209 00:17:55,330 --> 00:18:01,990 acceptance of the proposed model of reality as gospel. He says fortunately few 210 00:18:01,990 --> 00:18:05,390 would argue that all relations between people can be quantified and manipulated 211 00:18:05,390 --> 00:18:09,559 by formulas of human behavior, no faultless assumptions and so no perfect 212 00:18:09,559 --> 00:18:14,679 model can be made. Tim Harford also refers to faith when he describes four 213 00:18:14,679 --> 00:18:19,980 assumptions underlying Big Data research, the first uncanny accuracy is easy to 214 00:18:19,980 --> 00:18:24,299 overrate, if we simply ignore false positives, oh sorry, the claim that 215 00:18:24,299 --> 00:18:27,320 causation has been knocked off its pedestal is fine if we are making 216 00:18:27,320 --> 00:18:31,539 predictions in the stable environment, but not if the world is changing. If you 217 00:18:31,539 --> 00:18:35,000 do not understand why things correlate, you cannot know what might breakdown this 218 00:18:35,000 --> 00:18:38,620 correlation either. The promise that sampling bias does not 219 00:18:38,620 --> 00:18:43,320 matter in such large data sets is simply not true, there is lots of bias in data 220 00:18:43,320 --> 00:18:47,950 sets, as for the idea of why with enough data, the numbers speak for themselves 221 00:18:47,950 --> 00:18:53,720 that seems hopelessly naive, in data sets where spurious patterns vastly outnumber 222 00:18:53,720 --> 00:18:58,299 genuine discoveries. This last point is described by Nicholas Taleb who writes 223 00:18:58,299 --> 00:19:02,700 that big data research has brought cherry- picking to an industrial level. Liam Weber 224 00:19:02,700 --> 00:19:08,140 in a 2007 paper demonstrated that data mining techniques could show a strong, but 225 00:19:08,140 --> 00:19:13,549 spurious relationship between the changes in the S&P 500 stock index and butter 226 00:19:13,549 --> 00:19:18,740 production in Bangladesh. What is strange about this mythology, that large data sets 227 00:19:18,740 --> 00:19:23,419 offer some higher form of intelligences, is that is paradoxical it attributes human 228 00:19:23,419 --> 00:19:26,730 qualities to something, while at the same time considering it to be more objective 229 00:19:26,730 --> 00:19:32,070 and more accurate than humans, but these beliefs can exist side by side. Consulting 230 00:19:32,070 --> 00:19:36,259 this Oracle and critically has quite far- reaching implications. 231 00:19:36,259 --> 00:19:40,490 For one it dehumanizes humans by asserting that human involvement through hypothesis 232 00:19:40,490 --> 00:19:46,950 and interpretation is unreliable and only by removing ourselves from the equation, 233 00:19:46,950 --> 00:19:51,100 can we finally see the world as it is. The practical consequence of this dynamic 234 00:19:51,100 --> 00:19:55,450 is that it is no longer possible to argue with the outcome of big data analysis 235 00:19:55,450 --> 00:19:59,770 because first of all it's supposedly bias free, interpretation free, you can't 236 00:19:59,770 --> 00:20:04,210 question it, you cannot check if it is bias free because the algorithms governing 237 00:20:04,210 --> 00:20:09,160 the analysis are often completely opaque. This becomes painful when you find 238 00:20:09,160 --> 00:20:13,159 yourself in the wrong category of a social sorting algorithm guiding real-world 239 00:20:13,159 --> 00:20:18,280 decisions on insurance, mortgage, work border check, scholarships and so on. 240 00:20:18,280 --> 00:20:23,919 Exclusion from certain privileges is only the most optimistic scenario, so it is not 241 00:20:23,919 --> 00:20:28,179 as effective as we might hope. It has a dehumanizing dark side 242 00:20:28,179 --> 00:20:30,910 so why do we believe. How did we come so infatuated 243 00:20:30,910 --> 00:20:35,200 with information. Our idea about information changed radically in the 244 00:20:35,200 --> 00:20:39,830 previous century from small statement of fact, to the essence of man's inner life 245 00:20:39,830 --> 00:20:43,870 and this shift started with the advent of cybernetics and information theory in the 246 00:20:43,870 --> 00:20:48,200 40s and 50s where information was suddenly seen as a means to control a system, any 247 00:20:48,200 --> 00:20:53,429 system be it mechanical physical, biological, cognitive or social. Here you 248 00:20:53,429 --> 00:20:58,410 see Norbert Wiener's moths a machine he built as part of a public relations stunt 249 00:20:58,410 --> 00:21:03,370 financed by Life magazine. The photos with him and his moth were unfortunately never 250 00:21:03,370 --> 00:21:07,470 published, because according to Life's editors, it didn't illustrate the human 251 00:21:07,470 --> 00:21:12,540 characteristics of computers very well. Norbert Wiener in the human hues of human 252 00:21:12,540 --> 00:21:15,090 beings wrote, "to live effectively is to live with 253 00:21:15,090 --> 00:21:19,340 adequate information, thus communication and control belong to the essence of man's 254 00:21:19,340 --> 00:21:23,539 inner life, even as they belong to his life in society" and almost 255 00:21:23,539 --> 00:21:27,570 simultaneously, Shannon published a mathematical theory of communication a 256 00:21:27,570 --> 00:21:31,720 theory of signals transmitted over distance. John Durham Peters in speaking 257 00:21:31,720 --> 00:21:36,389 into the air, describes how over time this information theory got reinterpreted by 258 00:21:36,389 --> 00:21:41,040 social scientists who mistook signal for significance. 259 00:21:41,040 --> 00:21:45,299 Or at Halpern in beautiful data describes how Alan Turing and Bertrand Russell had 260 00:21:45,299 --> 00:21:49,659 proved conclusively in struggling with the Entscheidungsproblem that many analytic 261 00:21:49,659 --> 00:21:54,090 functions could not be logically represented or mechanically executed and 262 00:21:54,090 --> 00:21:58,470 therefore machines were not human minds. She asks the very important question of 263 00:21:58,470 --> 00:22:03,320 why we have forgotten this history and do we still regularly equate reason with 264 00:22:03,320 --> 00:22:08,031 rationality. Having forgotten this ten years later in '58, artificial 265 00:22:08,031 --> 00:22:11,929 intelligence research began comparing computers and humans. Simon and Newell 266 00:22:11,929 --> 00:22:15,740 wrote: the programmed computer and human problem solver are both species belonging 267 00:22:15,740 --> 00:22:19,220 to the genus 'information processing system'. 268 00:22:19,220 --> 00:22:23,390 In the 80s, information was granted an even more powerful status: that of 269 00:22:23,390 --> 00:22:27,669 commodity. Like it or not, information has finally surpassed material goods as our 270 00:22:27,669 --> 00:22:41,009 basic resource. Bon appetit! How did we become so infatuated with information? 271 00:22:41,009 --> 00:22:50,890 Hey sorry *sighs* yeah, this is an image of a medieval drawing where the humors, 272 00:22:50,890 --> 00:22:57,009 the liquids in the body were seen as the the essence of our intelligence in the 273 00:22:57,009 --> 00:23:02,610 functioning of our system. A metaphor for our intelligence by the 1500s automata 274 00:23:02,610 --> 00:23:06,190 powered by Springs and gears had been devised, inspiring leading thinkers such 275 00:23:06,190 --> 00:23:11,669 as Rene Descartes to assert that humans are complex machines. The mind or soul was 276 00:23:11,669 --> 00:23:15,190 immaterial, completely separated from the body - only able to interact with the body 277 00:23:15,190 --> 00:23:18,909 through the pineal gland, which he considered the seat of the soul. 278 00:23:18,909 --> 00:23:22,760 And we still do it, the brain is commonly compared to a computer with the role of 279 00:23:22,760 --> 00:23:26,809 physical hardware played by the brain, and our thoughts serving a software. The brain 280 00:23:26,809 --> 00:23:31,769 is information processor. It is a metaphor that is sometimes mistaken for reality. 281 00:23:31,769 --> 00:23:36,150 Because of this the belief in the Oracle of big data is not such a great leap. 282 00:23:36,150 --> 00:23:39,440 Information is the essence of consciousness in this view. We've come 283 00:23:39,440 --> 00:23:45,169 full circle, we see machines as human like and view ourselves as machines. So does it 284 00:23:45,169 --> 00:23:48,309 work, we started out with two survival strategies predicting the behavior of 285 00:23:48,309 --> 00:23:52,090 others through anthropomorphizing and trying to predict the future through 286 00:23:52,090 --> 00:23:55,710 oracles. The first has helped us survive in the past, allows us to be empathic 287 00:23:55,710 --> 00:24:00,440 towards others - human and non-human. The second has comforted us throughout the 288 00:24:00,440 --> 00:24:03,769 ages, creating the idea of control of being able to predict and prevent 289 00:24:03,769 --> 00:24:07,809 disaster. So how are they working for us today? 290 00:24:07,809 --> 00:24:11,809 We definitely have reasons to be concerned with the sword of Damocles hanging over 291 00:24:11,809 --> 00:24:16,260 our heads: global warming setting in motion a chain of catastrophes threatening 292 00:24:16,260 --> 00:24:20,910 our survival, facing the inevitable death of capitalism's myth of eternal growth as 293 00:24:20,910 --> 00:24:26,970 Earth's research has run out we are in a bit of a pickle. Seeing our consciousness 294 00:24:26,970 --> 00:24:31,210 as separate from our bodies, like software and hardware. That offers some comforting 295 00:24:31,210 --> 00:24:34,190 options. One option is that since human 296 00:24:34,190 --> 00:24:37,820 consciousness is so similar to computer software, it can be transferred to a 297 00:24:37,820 --> 00:24:42,940 computer. Ray Kurzweil for example believes that it will soon be possible to 298 00:24:42,940 --> 00:24:48,160 download human minds to a computer, with immortality as a result. "Alliance to 299 00:24:48,160 --> 00:24:52,279 Rescue Civilization" by Burrows and Shapiro is a project that aims to back up 300 00:24:52,279 --> 00:24:56,110 human civilization in a lunar facility. The project artificially separates the 301 00:24:56,110 --> 00:25:01,000 hardware of the planet with its oceans and soils, and a data of human civilization. 302 00:25:01,000 --> 00:25:04,289 And last but not least, the most explicit and radical separation as well as the 303 00:25:04,289 --> 00:25:08,609 least optimistic outlook on our future, Elon Musk's SpaceX planned to colonize 304 00:25:08,609 --> 00:25:13,009 Mars, presented in September last year. The goal of the presentation being to make 305 00:25:13,009 --> 00:25:18,470 living on Mars seemed possible within our lifetime. Possible - and fun. 306 00:25:18,470 --> 00:25:22,460 A less extreme version of these attempts to escape doom is what, that with so much 307 00:25:22,460 --> 00:25:26,409 data at our fingertips and clever scientists, will figure out a way to solve 308 00:25:26,409 --> 00:25:30,419 our problems. Soon we'll laugh at our panic over global warming safely aboard 309 00:25:30,419 --> 00:25:34,790 our CO2 vacuum cleaners. With this belief we don't have to change our lives, our 310 00:25:34,790 --> 00:25:40,360 economies, our politics. We can carry on without making radical changes. Is this 311 00:25:40,360 --> 00:25:46,239 apathy warranted? What is happening while we are filling up the world's hard disks? 312 00:25:46,239 --> 00:25:50,009 Well, information is never disembodied, it always needs a carrier and the minerals 313 00:25:50,009 --> 00:25:54,019 used in the technology hosting our data come from conflict zones, resulting in 314 00:25:54,019 --> 00:25:58,549 slavery and ecocide. As for instance in the coltan and cassiterite mines in Congo, 315 00:25:58,549 --> 00:26:02,679 gold mines in Ghana. Minerals used in technology hosting our data come from 316 00:26:02,679 --> 00:26:07,399 unregulated zones leading to extreme pollution, as here in the black sludge 317 00:26:07,399 --> 00:26:15,850 lake in Baotou in China. EU waste is exported to unregulated zones, and server 318 00:26:15,850 --> 00:26:20,070 farms spit out an equal amount of CO2 as the global aviation industry. Our data 319 00:26:20,070 --> 00:26:23,860 cannot be separated from the physical, and its physical side is not so pretty. 320 00:26:23,860 --> 00:26:27,789 And what is happening is that the earth is getting warmer and climate research is not 321 00:26:27,789 --> 00:26:32,039 based on Twitter feeds, but our measurements yet somehow largely has been 322 00:26:32,039 --> 00:26:36,539 ignored for decades. Scientific consensus was reached in the 80s, and if you compare 323 00:26:36,539 --> 00:26:40,070 the dangerously slow response to this, to the response given to the threat of 324 00:26:40,070 --> 00:26:44,789 terrorism which has rapidly led to new laws, even new presidents, this shows how 325 00:26:44,789 --> 00:26:48,799 stories, metaphors, and mythologies in the world of social beings have more impact 326 00:26:48,799 --> 00:26:52,740 than scientific facts. And how threats that require drastic changes to the status 327 00:26:52,740 --> 00:26:59,730 quo are willfully ignored. So does this survival strategy work? This 328 00:26:59,730 --> 00:27:02,990 mythology, this belief in taking ourselves out of the equation, to bring us closer to 329 00:27:02,990 --> 00:27:07,260 truth, to reality as it is, separating ourselves from that which we observe, 330 00:27:07,260 --> 00:27:12,590 blinds us to the trouble we are in. And our true nature and embodied intelligence, 331 00:27:12,590 --> 00:27:17,620 not a brain in a jar, an organism completely intertwined with its 332 00:27:17,620 --> 00:27:21,240 environment, its existence completely dependent on the survival of the organisms 333 00:27:21,240 --> 00:27:25,679 it shares this planet with, we can't help to anthropomorphize, to approach 334 00:27:25,679 --> 00:27:30,390 everything around us as part of our social sphere with minds and agencies. And that 335 00:27:30,390 --> 00:27:35,320 is fine, it makes us human. It allows us to study the world around us with empathy. 336 00:27:35,320 --> 00:27:39,090 The most important thing is that the metaphor is not mistaken for reality. The 337 00:27:39,090 --> 00:27:45,009 computer creating, thinking, memorizing, writing, reading, learning, understanding, 338 00:27:45,009 --> 00:27:50,169 and people being hard-wired, stuck in a loop, unable to compute, interfacing with, 339 00:27:50,169 --> 00:27:55,710 and reprogramming ourselves - those metaphors are so embedded in our culture. 340 00:27:55,710 --> 00:27:59,159 You can only hope to create awareness about them. If there is more awareness 341 00:27:59,159 --> 00:28:02,480 about the misleading descriptions of machines as human-like and humans as 342 00:28:02,480 --> 00:28:06,820 machine-like and all of reality as an information process, it is more likely 343 00:28:06,820 --> 00:28:10,139 that there will be less blind enchantment with certain technology, and more 344 00:28:10,139 --> 00:28:14,249 questions asked about its purpose and demands. 345 00:28:14,249 --> 00:28:18,980 There is no strong AI... yet, only very clever automation. At this moment in 346 00:28:18,980 --> 00:28:23,419 history machines are proxies for human agendas and ideologies. There are many 347 00:28:23,419 --> 00:28:32,679 issues that need addressing. As Kate Crawford and Meredith Whittaker point out 348 00:28:32,679 --> 00:28:37,369 in the AI Now report, recent examples of AI deployments such as during the US 349 00:28:37,369 --> 00:28:41,190 elections and Brexit, or Facebook revealing teenagers emotional states to 350 00:28:41,190 --> 00:28:45,539 advertisers looking to target depressed teens, show how the interests of those 351 00:28:45,539 --> 00:28:49,039 deploying advanced data systems can overshadow the public interest, acting in 352 00:28:49,039 --> 00:28:53,080 ways contrary to individual autonomy and collective welfare, often without this 353 00:28:53,080 --> 00:28:58,630 being visible at all to those affected. The report points to many - I highly 354 00:28:58,630 --> 00:29:04,769 recommend reading it - and here are a few concerns. Concerns about social safety 355 00:29:04,769 --> 00:29:08,290 nets and human resource distributions when the dynamic of labor and employment 356 00:29:08,290 --> 00:29:12,691 change. Workers most likely to be affected are women and minorities. Automated 357 00:29:12,691 --> 00:29:16,899 decision-making systems are often unseen and there are few established means to 358 00:29:16,899 --> 00:29:20,460 assess their fairness, to contest and rectify wrong or harmful decisions or 359 00:29:20,460 --> 00:29:29,590 impacts. Those directly impacted.... Sorry, 360 00:29:29,590 --> 00:29:35,570 automated... No, sorry.... I'm lost... Those directly impacted by deployment of 361 00:29:35,570 --> 00:29:49,909 AI systems rarely have a role in designing them. *sighs* And to assess their 362 00:29:49,909 --> 00:29:54,070 fairness, to confess and rectify wrong and harmful decisions or impacts, lacks.... 363 00:29:54,070 --> 00:29:58,960 lack of methods measuring and assessing social and economic impacts... nah, let's 364 00:29:58,960 --> 00:30:09,240 keep scrolling back.... In any case, there is a great chance of like me bias because 365 00:30:09,240 --> 00:30:17,299 of the uniform... uniformity of those developing these systems. Seeing the 366 00:30:17,299 --> 00:30:20,889 Oracle we've constructed for what it is means to stop comforting, comforting 367 00:30:20,889 --> 00:30:25,269 ourselves, to ask questions. A quote from super intelligence, the idea that it's 368 00:30:25,269 --> 00:30:29,800 smart people by (?)Muchaichai(?) (?)Clowsky(?), the pressing ethical questions in machine 369 00:30:29,800 --> 00:30:33,430 learning are not about machines becoming self-aware and taking over the world, but 370 00:30:33,430 --> 00:30:37,910 about how people can exploit other people. Or through careless, in carelessness 371 00:30:37,910 --> 00:30:43,070 introduce immoral behavior into automated systems. Instead of waiting for the nerd 372 00:30:43,070 --> 00:30:46,860 rapture, or for Elon Musk to whisk us off the planet, it is important to come to 373 00:30:46,860 --> 00:30:50,750 terms with a more modest perception of ourselves and our machines. Facing the 374 00:30:50,750 --> 00:30:56,590 ethical repercussions of the systems we are putting in place. Having the real 375 00:30:56,590 --> 00:31:02,320 discussion, not the one we hope for, but the hard one that requires actual change 376 00:31:02,320 --> 00:31:06,540 and a new mythology. One that works, not only for us, but for all those human and 377 00:31:06,540 --> 00:31:11,800 non-human, we share the planet with. Thank you. That's it. 378 00:31:11,800 --> 00:31:24,170 *applause* 379 00:31:24,170 --> 00:31:30,209 Herald Angel: Thank you, Marloes. Is there any questions? Like, you would have one 380 00:31:30,209 --> 00:31:36,689 minute. *laugs* Okay. So, thank you again. Give her a big applause again, 381 00:31:36,689 --> 00:31:39,469 thank you. 382 00:31:39,469 --> 00:31:43,976 *applause* 383 00:31:43,976 --> 00:31:49,958 *34c3 outro* 384 00:31:49,958 --> 00:32:04,901 subtitles created by c3subtitles.de in the year 2018. Join, and help us!