0 00:00:00,000 --> 00:00:30,000 Dear viewer, these subtitles were generated by a machine via the service Trint and therefore are (very) buggy. If you are capable, please help us to create good quality subtitles: https://c3subtitles.de/talk/734 Thanks! 1 00:00:14,990 --> 00:00:17,119 Our next presentation comes from a 2 00:00:17,120 --> 00:00:19,459 man who is from Great Britain 3 00:00:19,460 --> 00:00:21,149 but now lives in Greece. 4 00:00:21,150 --> 00:00:23,209 Now, I don't know if you just try 5 00:00:23,210 --> 00:00:25,039 to escape the bad weather in Britain or 6 00:00:25,040 --> 00:00:27,119 the Brexit, but we're sure 7 00:00:27,120 --> 00:00:29,689 a happy he made all the way 8 00:00:29,690 --> 00:00:31,030 here to Hamburg today. 9 00:00:34,630 --> 00:00:37,149 Now, the latest elections in the US 10 00:00:37,150 --> 00:00:39,579 and the referendum on 11 00:00:39,580 --> 00:00:41,649 the Brexit have shown us that 12 00:00:41,650 --> 00:00:44,049 with all the collected data 13 00:00:44,050 --> 00:00:46,149 and technology that we use 14 00:00:46,150 --> 00:00:48,639 to predict the outcome of 15 00:00:48,640 --> 00:00:49,720 such elections, 16 00:00:50,950 --> 00:00:52,420 we can do it. We cannot 17 00:00:53,560 --> 00:00:55,479 really know everything or predict 18 00:00:55,480 --> 00:00:56,480 everything. 19 00:00:57,250 --> 00:01:00,189 And you dark age as an explanation 20 00:01:00,190 --> 00:01:02,379 of what we cannot no longer 21 00:01:02,380 --> 00:01:03,999 know about this world and what we can do. 22 00:01:04,000 --> 00:01:06,429 About this hour, our speaker, James, 23 00:01:06,430 --> 00:01:09,129 here is an artist, 24 00:01:09,130 --> 00:01:11,619 a publisher and a writer. 25 00:01:11,620 --> 00:01:13,689 And his writings have appeared in 26 00:01:13,690 --> 00:01:15,549 magazines and newspapers all around the 27 00:01:15,550 --> 00:01:18,009 world, such as Wired, 28 00:01:18,010 --> 00:01:20,799 The Observer and The Guardian. 29 00:01:20,800 --> 00:01:23,139 He's also a very experienced 30 00:01:23,140 --> 00:01:25,359 lecturer in 31 00:01:25,360 --> 00:01:27,519 conferences, in universities 32 00:01:27,520 --> 00:01:29,389 and at different events. 33 00:01:29,390 --> 00:01:31,239 So we're very happy he's here today. 34 00:01:31,240 --> 00:01:33,459 So please welcome him and give him a warm 35 00:01:33,460 --> 00:01:34,480 round of applause. 36 00:01:35,800 --> 00:01:36,800 James Freital. 37 00:01:44,500 --> 00:01:46,569 Hello, uh, thank you very much for that 38 00:01:46,570 --> 00:01:48,579 introduction. Uh, thank you very much for 39 00:01:48,580 --> 00:01:49,509 having me. 40 00:01:49,510 --> 00:01:50,889 Thank you very much for coming here. 41 00:01:50,890 --> 00:01:52,539 This is huge and amazing and I'm very 42 00:01:52,540 --> 00:01:54,940 grateful. And I was, um. 43 00:01:56,450 --> 00:01:58,489 Yeah, I just put hello, there are the new 44 00:01:58,490 --> 00:02:00,109 Dark Age because I didn't want it to 45 00:02:00,110 --> 00:02:02,329 sound quite so frightening at the outset. 46 00:02:02,330 --> 00:02:04,549 I don't know, like it's a thing 47 00:02:04,550 --> 00:02:05,929 I'm working on at the moment. 48 00:02:05,930 --> 00:02:08,149 And I was also asked today to talk a bit 49 00:02:08,150 --> 00:02:10,219 broadly about some of the work that I do. 50 00:02:10,220 --> 00:02:11,539 So a lot of this is going to be like a 51 00:02:11,540 --> 00:02:13,189 bit me, me, me, and that I'm going to 52 00:02:13,190 --> 00:02:15,439 show some things 53 00:02:15,440 --> 00:02:17,749 I've made as a way of explaining 54 00:02:17,750 --> 00:02:20,149 the way that I think through technology 55 00:02:20,150 --> 00:02:21,529 and society and various stuff around 56 00:02:21,530 --> 00:02:23,419 that. And hopefully there'll be 57 00:02:23,420 --> 00:02:24,619 connections to that. 58 00:02:24,620 --> 00:02:26,689 And I'll talk quite a lot about 59 00:02:26,690 --> 00:02:28,819 one particular project I did this 60 00:02:28,820 --> 00:02:30,889 year and then 61 00:02:30,890 --> 00:02:33,649 some kind of extrapolations from that 62 00:02:33,650 --> 00:02:35,719 to kind of where I think I 63 00:02:35,720 --> 00:02:37,849 and perhaps we need to be thinking 64 00:02:37,850 --> 00:02:38,599 next. 65 00:02:38,600 --> 00:02:41,059 And I actually my background 66 00:02:41,060 --> 00:02:43,039 is as a computer scientist, but I'm a 67 00:02:43,040 --> 00:02:44,899 seriously lapsed one. 68 00:02:44,900 --> 00:02:46,339 I managed to graduate with a masters 69 00:02:46,340 --> 00:02:47,389 degree without actually learning to 70 00:02:47,390 --> 00:02:49,309 program, which tells you everything you 71 00:02:49,310 --> 00:02:50,899 need to know about sex education in the 72 00:02:50,900 --> 00:02:53,179 UK and has since been trying 73 00:02:53,180 --> 00:02:55,369 to teach myself desperately ever since 74 00:02:55,370 --> 00:02:56,479 and really badly. 75 00:02:56,480 --> 00:02:58,579 But I've also been a literary publisher 76 00:02:58,580 --> 00:03:00,139 and a journalist in a number of other 77 00:03:00,140 --> 00:03:01,069 things. 78 00:03:01,070 --> 00:03:02,809 Um, but it was my kind of background in 79 00:03:02,810 --> 00:03:04,909 publishing and my interest in the Web 80 00:03:04,910 --> 00:03:06,979 that sort of got me slowly moving 81 00:03:06,980 --> 00:03:09,169 towards the place I made now, which 82 00:03:09,170 --> 00:03:11,089 is increasingly in the art world, but 83 00:03:11,090 --> 00:03:12,769 always, always crossing over. 84 00:03:12,770 --> 00:03:15,049 And this is a 85 00:03:15,050 --> 00:03:17,239 good example of the kind of where 86 00:03:17,240 --> 00:03:19,069 I put things together, which is a few 87 00:03:19,070 --> 00:03:20,070 years old now. 88 00:03:21,290 --> 00:03:23,029 But it was when I was working as a 89 00:03:23,030 --> 00:03:24,859 publisher, but wanted to start to talk to 90 00:03:24,860 --> 00:03:27,349 people about how technology 91 00:03:27,350 --> 00:03:28,849 and the Internet in particular changed 92 00:03:28,850 --> 00:03:30,649 the way in which we putting knowledge 93 00:03:30,650 --> 00:03:31,819 together. 94 00:03:31,820 --> 00:03:34,099 And so what I did was I took 95 00:03:34,100 --> 00:03:36,259 the complete history of a single 96 00:03:36,260 --> 00:03:37,189 Wikipedia article. 97 00:03:37,190 --> 00:03:39,439 I took the entire changelog of it, which 98 00:03:39,440 --> 00:03:41,209 was the article on the Iraq war, which at 99 00:03:41,210 --> 00:03:42,409 the time I made this book 100 00:03:44,060 --> 00:03:46,549 was it was about seven years of data 101 00:03:46,550 --> 00:03:48,349 on a single Wikipedia article. 102 00:03:48,350 --> 00:03:49,879 And because I was working at a publisher 103 00:03:49,880 --> 00:03:51,379 at the time, I learned how to make books 104 00:03:51,380 --> 00:03:53,359 properly. So I typeset it all. 105 00:03:53,360 --> 00:03:54,799 It's all in like neat little three 106 00:03:54,800 --> 00:03:57,379 columns, quite small type. 107 00:03:57,380 --> 00:03:59,539 But when I had that all print and bound 108 00:03:59,540 --> 00:04:01,819 that single Wikipedia article, Changelog 109 00:04:01,820 --> 00:04:04,669 took a 12 volume size 110 00:04:04,670 --> 00:04:06,949 so that the scale of an 111 00:04:06,950 --> 00:04:09,519 awful old style encyclopedia 112 00:04:09,520 --> 00:04:11,659 and what I was thinking through this work 113 00:04:11,660 --> 00:04:12,660 is the fact that 114 00:04:14,060 --> 00:04:15,060 the way in which 115 00:04:16,279 --> 00:04:18,409 we build stuff with technology 116 00:04:18,410 --> 00:04:20,838 doesn't really necessarily change 117 00:04:20,839 --> 00:04:22,069 the world so often. But the way we 118 00:04:22,070 --> 00:04:24,559 understand how that technology functions 119 00:04:24,560 --> 00:04:26,299 changes the way in which we can interact 120 00:04:26,300 --> 00:04:27,229 with the world. 121 00:04:27,230 --> 00:04:29,299 And so this is still an old school 122 00:04:29,300 --> 00:04:30,559 encyclopedia. 123 00:04:30,560 --> 00:04:32,149 Wikipedia is still an old style 124 00:04:32,150 --> 00:04:34,099 encyclopedia, but the technology we've 125 00:04:34,100 --> 00:04:36,229 built that underlies it allows 126 00:04:36,230 --> 00:04:37,459 us to put that together in a way that 127 00:04:37,460 --> 00:04:39,319 allows us to see all the changes we made 128 00:04:39,320 --> 00:04:40,789 to it. We can see the fact that 129 00:04:40,790 --> 00:04:43,309 encyclopedias are not final 130 00:04:43,310 --> 00:04:45,469 statements of fact, but processes 131 00:04:45,470 --> 00:04:47,779 of argument over time with many, many, 132 00:04:47,780 --> 00:04:49,519 many voices contributing to them. 133 00:04:49,520 --> 00:04:51,709 It's a machine for historiography rather 134 00:04:51,710 --> 00:04:52,729 than for history. 135 00:04:52,730 --> 00:04:53,929 And that, for me, captures something 136 00:04:53,930 --> 00:04:55,729 quite fundamentally different about the 137 00:04:55,730 --> 00:04:56,989 way we're capable of building stuff 138 00:04:56,990 --> 00:04:58,039 today. 139 00:04:58,040 --> 00:05:00,049 And so my projects go back and forth 140 00:05:00,050 --> 00:05:02,209 between like things that exist as 141 00:05:02,210 --> 00:05:04,339 big, solid objects, which are kind 142 00:05:04,340 --> 00:05:06,409 of quite easy to understand 143 00:05:06,410 --> 00:05:08,359 and more kind of material software 144 00:05:08,360 --> 00:05:10,519 things. And this was 145 00:05:10,520 --> 00:05:12,739 a project that I ran for three and a half 146 00:05:12,740 --> 00:05:14,989 years on various social media 147 00:05:14,990 --> 00:05:17,119 tracking reports of drone strikes. 148 00:05:17,120 --> 00:05:19,469 So I was working with journalists 149 00:05:19,470 --> 00:05:20,719 from places like the Bureau of Investigative 150 00:05:20,720 --> 00:05:22,759 Journalism who were tracking drone 151 00:05:22,760 --> 00:05:23,760 strikes in 152 00:05:25,040 --> 00:05:26,629 undeclared theaters of war. 153 00:05:26,630 --> 00:05:28,759 So these were the covert ones in Pakistan 154 00:05:28,760 --> 00:05:30,499 and Yemen and Somalia. 155 00:05:30,500 --> 00:05:32,659 These reports were being posted and 156 00:05:32,660 --> 00:05:34,129 those who are kind of interested in them 157 00:05:34,130 --> 00:05:36,019 could find out slightly more about them. 158 00:05:36,020 --> 00:05:37,819 But it struck me as deeply strange that 159 00:05:37,820 --> 00:05:39,959 this entire war was going on 160 00:05:39,960 --> 00:05:42,229 with our images, which 161 00:05:42,230 --> 00:05:44,209 is deeply strange in our current times. 162 00:05:44,210 --> 00:05:45,919 I mean, and for a long time, for more 163 00:05:45,920 --> 00:05:47,749 than a century, we've they used to send 164 00:05:47,750 --> 00:05:49,369 illustrators to battlefields, write with 165 00:05:49,370 --> 00:05:50,689 pencils and papers. 166 00:05:50,690 --> 00:05:51,859 In an age of mass media where we're 167 00:05:51,860 --> 00:05:53,869 saturated with images, there was this 168 00:05:53,870 --> 00:05:56,029 whole war going on without them. 169 00:05:56,030 --> 00:05:57,469 And yet at the same time, we spent the 170 00:05:57,470 --> 00:06:00,169 last 10, 20 years photographing 171 00:06:00,170 --> 00:06:01,909 the entire earth from space. 172 00:06:01,910 --> 00:06:04,309 You can pull out your phone and look into 173 00:06:04,310 --> 00:06:06,559 any spot on the Earth's surface. 174 00:06:06,560 --> 00:06:08,389 And I wanted to pull those kind of 175 00:06:08,390 --> 00:06:10,429 infrastructures together to close those 176 00:06:10,430 --> 00:06:12,319 loops. So every time there was one of 177 00:06:12,320 --> 00:06:13,939 these covert drone strikes reported, I 178 00:06:13,940 --> 00:06:16,279 went to a digital mapping service, 179 00:06:16,280 --> 00:06:18,409 use various ones to find as 180 00:06:18,410 --> 00:06:20,479 close as I could the actual landscape 181 00:06:20,480 --> 00:06:21,589 in which that occurred. 182 00:06:21,590 --> 00:06:23,569 It's kind of re-supply some of these 183 00:06:23,570 --> 00:06:25,669 images back to this ongoing 184 00:06:25,670 --> 00:06:26,670 fight. 185 00:06:27,250 --> 00:06:29,029 And I post those to Instagram, to 186 00:06:29,030 --> 00:06:30,379 Twitter, to Tumblr is these long 187 00:06:30,380 --> 00:06:32,479 sequences and trying to create 188 00:06:32,480 --> 00:06:34,249 a kind of visual record of what was 189 00:06:34,250 --> 00:06:36,049 occurring. One of the good things that 190 00:06:36,050 --> 00:06:38,509 happened is that these images started 191 00:06:38,510 --> 00:06:40,459 to then be used in other places. 192 00:06:40,460 --> 00:06:42,949 So media, when discussing drone strikes, 193 00:06:42,950 --> 00:06:45,079 instead of just running like a photo of 194 00:06:45,080 --> 00:06:47,149 a Predator drone supplied to them 195 00:06:47,150 --> 00:06:48,949 by the US military would actually show 196 00:06:48,950 --> 00:06:50,629 images of the places where these things 197 00:06:50,630 --> 00:06:52,909 were occurring, the actual landscapes 198 00:06:52,910 --> 00:06:54,259 that we're kind of talking about here. 199 00:06:55,390 --> 00:06:58,089 And that attempt to make these kind of 200 00:06:58,090 --> 00:06:59,559 hard to see, I don't want to say 201 00:06:59,560 --> 00:07:01,779 invisible because you can always say 202 00:07:01,780 --> 00:07:03,909 thanks and good and I 203 00:07:03,910 --> 00:07:05,499 don't want to say that, but it is hard to 204 00:07:05,500 --> 00:07:06,849 see things kind of more visible. 205 00:07:06,850 --> 00:07:08,409 I take that kind of out into the real 206 00:07:08,410 --> 00:07:09,519 world as well. 207 00:07:09,520 --> 00:07:10,929 So for several years, I've been doing 208 00:07:10,930 --> 00:07:12,189 these drone shadow works, 209 00:07:13,900 --> 00:07:16,539 which are one to one realizations. 210 00:07:16,540 --> 00:07:18,489 And that came of me just wanting to know 211 00:07:18,490 --> 00:07:20,409 more about these big machines, I think, 212 00:07:20,410 --> 00:07:22,359 which I am very unlikely to ever see in 213 00:07:22,360 --> 00:07:23,349 the real world. 214 00:07:23,350 --> 00:07:25,419 But I have, you know, the tools 215 00:07:25,420 --> 00:07:27,699 to find out just the size of it by 216 00:07:27,700 --> 00:07:29,619 just a simple thing like that and go and 217 00:07:29,620 --> 00:07:31,269 put that out in the street where people 218 00:07:31,270 --> 00:07:32,709 can actually see and kind of measure 219 00:07:32,710 --> 00:07:34,569 oneself physically against the thing. 220 00:07:34,570 --> 00:07:36,729 It really changes the way people see 221 00:07:36,730 --> 00:07:37,989 and understand these things. 222 00:07:37,990 --> 00:07:40,119 And I've painted these all over 223 00:07:40,120 --> 00:07:41,879 the world and it's totally open source 224 00:07:41,880 --> 00:07:43,989 that. So there's a guide online to join 225 00:07:43,990 --> 00:07:45,519 your own. And people have then 226 00:07:45,520 --> 00:07:47,619 subsequently taken that and done it both 227 00:07:47,620 --> 00:07:49,719 as art installations and 228 00:07:49,720 --> 00:07:51,099 protests and all these kind of. 229 00:07:51,100 --> 00:07:52,509 And I just get like people sending me 230 00:07:52,510 --> 00:07:53,809 pictures occasionally when these things 231 00:07:53,810 --> 00:07:55,869 have appeared around the world, which is 232 00:07:55,870 --> 00:07:57,099 amazing. 233 00:07:57,100 --> 00:07:59,289 It's very simple to just set out 234 00:07:59,290 --> 00:08:01,269 to do a thing in order to understand it 235 00:08:01,270 --> 00:08:03,699 and then realize a whole bunch of stuff. 236 00:08:04,970 --> 00:08:07,069 By doing that physical action, 237 00:08:07,070 --> 00:08:09,079 what the drone shadows will emphasize is 238 00:08:09,080 --> 00:08:11,299 actually the the intentional 239 00:08:11,300 --> 00:08:12,979 invisibility of these things, the fact 240 00:08:12,980 --> 00:08:15,229 that they're designed not to be seen, not 241 00:08:15,230 --> 00:08:17,449 to be seen physically because they fly 242 00:08:17,450 --> 00:08:19,129 so high, they're invisible, but not to be 243 00:08:19,130 --> 00:08:21,259 seen politically as well, because 244 00:08:21,260 --> 00:08:22,939 because they automate systems that would 245 00:08:22,940 --> 00:08:25,039 otherwise lead 246 00:08:25,040 --> 00:08:26,509 to body bags and all the reasons, you 247 00:08:26,510 --> 00:08:29,119 know, the military is like drones 248 00:08:29,120 --> 00:08:30,120 and. 249 00:08:31,250 --> 00:08:32,569 And that got me into the habit of doing 250 00:08:32,570 --> 00:08:34,579 other things in the street as well. 251 00:08:34,580 --> 00:08:36,678 This is a rainbow I drew 252 00:08:36,679 --> 00:08:38,449 in Ukraine a couple of years ago. 253 00:08:38,450 --> 00:08:40,579 And if you spend half as much 254 00:08:40,580 --> 00:08:42,649 time as I have on Google Maps, you 255 00:08:42,650 --> 00:08:44,479 might have seen the rainbow planes. 256 00:08:44,480 --> 00:08:46,219 That kind of I find them deeply 257 00:08:46,220 --> 00:08:48,319 beautiful. I find them if 258 00:08:48,320 --> 00:08:49,639 you're scouting around airports or 259 00:08:49,640 --> 00:08:51,589 whatever, you might spot these things in 260 00:08:51,590 --> 00:08:52,789 Google Maps and what they are. 261 00:08:52,790 --> 00:08:55,189 They're artifacts of satellite mapping. 262 00:08:55,190 --> 00:08:57,109 So the satellite is passing very fast 263 00:08:57,110 --> 00:08:59,269 overhead, taking images of the ground. 264 00:08:59,270 --> 00:09:01,219 The planes are flying very far below 265 00:09:01,220 --> 00:09:02,269 them. 266 00:09:02,270 --> 00:09:03,769 And the thing is that satellites don't 267 00:09:03,770 --> 00:09:05,989 see like humanzee. 268 00:09:05,990 --> 00:09:07,219 And this is the point at which I become 269 00:09:07,220 --> 00:09:08,449 fascinated because I'm interested in 270 00:09:08,450 --> 00:09:10,549 seeing how technology sees the 271 00:09:10,550 --> 00:09:12,529 world in fundamentally different ways 272 00:09:12,530 --> 00:09:13,819 than we do. 273 00:09:13,820 --> 00:09:15,919 And satellites see using a variety of 274 00:09:15,920 --> 00:09:18,199 sensors, both visible spectrum, 275 00:09:18,200 --> 00:09:20,599 red, green, blue and 276 00:09:20,600 --> 00:09:22,519 in a high definition, black and white in 277 00:09:22,520 --> 00:09:24,619 this case and deep into 278 00:09:24,620 --> 00:09:26,299 the infrared and the ultraviolet. 279 00:09:26,300 --> 00:09:27,919 So they see things that we're not capable 280 00:09:27,920 --> 00:09:30,919 of seeing. And this these little glitches 281 00:09:30,920 --> 00:09:32,719 are very interesting because they allow 282 00:09:32,720 --> 00:09:34,849 us to glimpse inside that system 283 00:09:34,850 --> 00:09:36,949 and see the way in which the machine 284 00:09:36,950 --> 00:09:39,199 is seeing the world and attempt to kind 285 00:09:39,200 --> 00:09:41,599 of reflect that back by 286 00:09:41,600 --> 00:09:43,669 making them visible and 287 00:09:43,670 --> 00:09:45,139 addressable and kind of pointing these 288 00:09:45,140 --> 00:09:46,729 things out. In this way, it makes it 289 00:09:46,730 --> 00:09:48,799 possible to to think about them 290 00:09:48,800 --> 00:09:51,259 from in slightly different ways. 291 00:09:51,260 --> 00:09:53,959 And a couple of years ago, 292 00:09:53,960 --> 00:09:56,479 I did this project in South London 293 00:09:56,480 --> 00:09:57,480 and. 294 00:09:59,730 --> 00:10:01,979 This was called the right to flight, and 295 00:10:01,980 --> 00:10:03,749 this is on top of a multistory car park 296 00:10:03,750 --> 00:10:05,279 in south London where 297 00:10:06,540 --> 00:10:08,609 we built these silos, one of which is a 298 00:10:08,610 --> 00:10:10,199 hangar for this blimp. 299 00:10:10,200 --> 00:10:11,909 And the blimp is kind of amazing. 300 00:10:11,910 --> 00:10:13,919 It's the thing called a helicopter, which 301 00:10:13,920 --> 00:10:16,019 is a hybrid of a balloon and a kite 302 00:10:16,020 --> 00:10:18,239 invented by slightly mad British guy who 303 00:10:18,240 --> 00:10:19,529 I had to spend several days in the pub 304 00:10:19,530 --> 00:10:21,869 with. And the helicopters 305 00:10:21,870 --> 00:10:23,759 are amazing because it's better than both 306 00:10:23,760 --> 00:10:25,979 balloon and it has helium to 307 00:10:25,980 --> 00:10:28,139 lift and but also a kite to 308 00:10:28,140 --> 00:10:29,549 fight the wind so it doesn't get blown 309 00:10:29,550 --> 00:10:31,079 down like a normal balloon, but has much 310 00:10:31,080 --> 00:10:32,759 more left than a regular kite. 311 00:10:32,760 --> 00:10:34,889 And and I was flying it 312 00:10:34,890 --> 00:10:37,679 off this rooftop to kind of investigate 313 00:10:37,680 --> 00:10:40,139 what it was like to be able to have 314 00:10:40,140 --> 00:10:42,149 that view from above, that kind of 315 00:10:42,150 --> 00:10:44,219 classic surveillant God's eye view 316 00:10:44,220 --> 00:10:45,749 of the world. How would it feel for me to 317 00:10:45,750 --> 00:10:47,729 be able to make a kind of aerial images 318 00:10:47,730 --> 00:10:51,059 like this to to become 319 00:10:51,060 --> 00:10:53,309 to to access this airspace 320 00:10:53,310 --> 00:10:55,379 above London that was is otherwise 321 00:10:55,380 --> 00:10:56,999 kind of heavily restricted. 322 00:10:57,000 --> 00:10:59,069 And and it was a beautiful project and 323 00:10:59,070 --> 00:11:00,269 it ran for several months. 324 00:11:00,270 --> 00:11:02,489 But it also waited me out quite a large 325 00:11:02,490 --> 00:11:04,559 extent because obviously 326 00:11:04,560 --> 00:11:05,879 the first thing I did was work a camera 327 00:11:05,880 --> 00:11:07,979 on this thing because aerial photos are 328 00:11:07,980 --> 00:11:09,389 beautiful and I want to see all this 329 00:11:09,390 --> 00:11:10,859 stuff from above. 330 00:11:10,860 --> 00:11:12,359 And very quickly, I started to become 331 00:11:12,360 --> 00:11:15,269 really uncomfortable with that because 332 00:11:15,270 --> 00:11:17,819 I was doing surveillance, essentially, 333 00:11:17,820 --> 00:11:20,009 and I was while I try to inform 334 00:11:20,010 --> 00:11:22,350 people what I was doing and I, 335 00:11:23,580 --> 00:11:25,259 I looked at various ways doing, I kind of 336 00:11:25,260 --> 00:11:26,909 released the footage openly so I would 337 00:11:26,910 --> 00:11:28,679 see the products of what I was doing in 338 00:11:28,680 --> 00:11:29,849 this kind of stuff. 339 00:11:29,850 --> 00:11:32,249 No one actually cared why 340 00:11:32,250 --> 00:11:33,869 this is in part Londoners not really 341 00:11:33,870 --> 00:11:36,029 giving a shit about anything, but also 342 00:11:38,100 --> 00:11:40,439 whether it's just a 343 00:11:40,440 --> 00:11:41,639 lack of awareness of the thing or whether 344 00:11:41,640 --> 00:11:43,229 it's just the fact that London is so 345 00:11:43,230 --> 00:11:44,729 subject to surveillance all the time 346 00:11:44,730 --> 00:11:47,039 anyway. And this is kind of off 347 00:11:47,040 --> 00:11:48,449 people's minds. But I was their 348 00:11:48,450 --> 00:11:50,759 responsibility was entirely on me 349 00:11:50,760 --> 00:11:53,279 to stop doing this, essentially 350 00:11:53,280 --> 00:11:54,509 that I suddenly got to a point where I 351 00:11:54,510 --> 00:11:57,179 couldn't make any more images like this. 352 00:11:57,180 --> 00:11:58,649 And I realized that in many ways it's 353 00:11:58,650 --> 00:12:00,029 quite hard to make artworks about 354 00:12:00,030 --> 00:12:01,589 surveillance that don't essentially do 355 00:12:01,590 --> 00:12:02,789 more surveillance. 356 00:12:02,790 --> 00:12:04,709 And that's a really bad pattern to get 357 00:12:04,710 --> 00:12:07,029 into because I don't think it helps. 358 00:12:07,030 --> 00:12:08,639 I think it just kind of reinforces the 359 00:12:08,640 --> 00:12:10,709 logics of surveillance and normalizes 360 00:12:10,710 --> 00:12:11,669 it for people. 361 00:12:11,670 --> 00:12:13,379 And so I stopped myself. 362 00:12:13,380 --> 00:12:15,269 I took the cameras off and I end up 363 00:12:15,270 --> 00:12:16,679 putting like a kind of pirate box on 364 00:12:16,680 --> 00:12:18,269 there so people could share that felt 365 00:12:18,270 --> 00:12:19,919 like a more useful 366 00:12:21,030 --> 00:12:23,009 kind of relationship to have with it. 367 00:12:23,010 --> 00:12:24,269 And so I start to look at more kind of 368 00:12:24,270 --> 00:12:26,069 pedagogical ways to talk about these 369 00:12:26,070 --> 00:12:28,919 things, like this project, 370 00:12:28,920 --> 00:12:31,259 which is called Citizen X, which 371 00:12:31,260 --> 00:12:34,529 is a browser extension. 372 00:12:34,530 --> 00:12:36,629 Very simply, anyone 373 00:12:36,630 --> 00:12:38,429 can download it tracks your browsing, but 374 00:12:38,430 --> 00:12:40,649 entirely privately data doesn't leave 375 00:12:40,650 --> 00:12:42,659 it. I do not share it with anyone, 376 00:12:42,660 --> 00:12:43,660 including myself. 377 00:12:44,730 --> 00:12:46,799 But as you're browsing the web, you can 378 00:12:46,800 --> 00:12:48,179 hit the button and it will pull down or 379 00:12:48,180 --> 00:12:49,949 show you very simply stuff. 380 00:12:49,950 --> 00:12:51,209 That's very simple for most people in 381 00:12:51,210 --> 00:12:53,579 this room, I'm sure where the Internet 382 00:12:53,580 --> 00:12:55,350 thinks you are and where 383 00:12:56,610 --> 00:12:57,779 the Internet thinks the website you're 384 00:12:57,780 --> 00:12:58,679 visiting is. 385 00:12:58,680 --> 00:13:00,839 So it shows you this relationship of the 386 00:13:00,840 --> 00:13:02,069 the physical infrastructure, the 387 00:13:02,070 --> 00:13:04,109 Internet, and like many artists, have 388 00:13:04,110 --> 00:13:05,699 been kind of investigating the physical 389 00:13:05,700 --> 00:13:06,689 infrastructure of the Internet for a 390 00:13:06,690 --> 00:13:08,759 while and thinking it's important to sort 391 00:13:08,760 --> 00:13:10,859 of point at this stuff and get this 392 00:13:10,860 --> 00:13:13,199 magic cloud place you hear about isn't 393 00:13:13,200 --> 00:13:14,669 some faraway, strange place. 394 00:13:14,670 --> 00:13:15,779 It's like it's a real thing. 395 00:13:15,780 --> 00:13:18,089 It's here. It's large buildings filled 396 00:13:18,090 --> 00:13:20,609 with computers in legal jurisdictions 397 00:13:20,610 --> 00:13:22,229 subject to certain laws. 398 00:13:22,230 --> 00:13:24,359 Um, but but there's there's 399 00:13:24,360 --> 00:13:26,549 weird stuff going on under that as well. 400 00:13:26,550 --> 00:13:28,319 So with this project, I didn't just want 401 00:13:28,320 --> 00:13:30,869 to highlight that physical Internet. 402 00:13:30,870 --> 00:13:32,369 I wanted to highlight what that physical 403 00:13:32,370 --> 00:13:34,469 Internet does to your rights, which 404 00:13:34,470 --> 00:13:36,899 is in this case based on NSA documents 405 00:13:36,900 --> 00:13:40,019 showing how the NSA assigns citizenship. 406 00:13:40,020 --> 00:13:42,299 And what the NSA does is it 407 00:13:42,300 --> 00:13:44,429 when it's pulling in all that data, it 408 00:13:44,430 --> 00:13:45,719 looks at your browsing habits and it 409 00:13:45,720 --> 00:13:47,909 says, well, based on where this person 410 00:13:47,910 --> 00:13:49,079 is going on the Internet and this sort of 411 00:13:49,080 --> 00:13:50,459 behavior they have, there's this 412 00:13:50,460 --> 00:13:52,829 percentage likelihood they're American 413 00:13:52,830 --> 00:13:54,299 because that's an important determinant 414 00:13:54,300 --> 00:13:56,879 for the NSA because they're supposed 415 00:13:56,880 --> 00:13:58,109 to not track Americans. 416 00:13:58,110 --> 00:14:00,329 So they have to have some mechanism for 417 00:14:00,330 --> 00:14:01,919 making this distinction. 418 00:14:01,920 --> 00:14:04,139 So moment by moment online, 419 00:14:04,140 --> 00:14:05,969 your citizenship, which used to be this 420 00:14:05,970 --> 00:14:07,709 really solid thing that used to be a 421 00:14:07,710 --> 00:14:09,149 thing determined by your passport, by 422 00:14:09,150 --> 00:14:11,789 your place of birth, that you could prove 423 00:14:11,790 --> 00:14:14,009 those of us who are lucky to hold stable 424 00:14:14,010 --> 00:14:15,629 citizenship. So you could prove with a 425 00:14:15,630 --> 00:14:17,699 document that is no longer the 426 00:14:17,700 --> 00:14:19,619 case, your right to privacy is being 427 00:14:19,620 --> 00:14:21,899 determined algorithmically all the time 428 00:14:21,900 --> 00:14:24,449 by tracking of your behavior. 429 00:14:24,450 --> 00:14:26,159 And so what's this next does is it looks 430 00:14:26,160 --> 00:14:28,439 at your and your 431 00:14:28,440 --> 00:14:30,959 very simple behaviors online, 432 00:14:30,960 --> 00:14:32,009 and it builds you one of these 433 00:14:32,010 --> 00:14:34,019 algorithmic citizenships to say this is 434 00:14:34,020 --> 00:14:36,389 what actually your web browsing behavior 435 00:14:36,390 --> 00:14:38,609 would look like if someone 436 00:14:38,610 --> 00:14:40,709 more nefarious was looking at it and how 437 00:14:40,710 --> 00:14:42,599 that might kind of come to affect your 438 00:14:42,600 --> 00:14:43,600 rights. 439 00:14:44,610 --> 00:14:46,829 And I was looking 440 00:14:46,830 --> 00:14:48,329 at the sky a while back because I was 441 00:14:48,330 --> 00:14:50,609 looking for these planes, which are Kovar 442 00:14:50,610 --> 00:14:52,859 Metropolitan Police surveillance 443 00:14:52,860 --> 00:14:54,929 planes, and hang out in fields quite 444 00:14:54,930 --> 00:14:57,029 a lot looking for weird stuff like this. 445 00:14:57,030 --> 00:14:58,269 And I'm. 446 00:14:58,270 --> 00:15:00,579 What I found instead was, was these 447 00:15:00,580 --> 00:15:02,739 flights, this 448 00:15:02,740 --> 00:15:04,329 is an example of where I go into fields 449 00:15:04,330 --> 00:15:06,969 and stare at the sky and find myself 450 00:15:06,970 --> 00:15:09,189 stood next to a bunch of kind of weird, 451 00:15:09,190 --> 00:15:11,349 interesting people, because you end up 452 00:15:11,350 --> 00:15:13,239 in the fields around airports with the 453 00:15:13,240 --> 00:15:15,459 plane spotters, the controllers 454 00:15:15,460 --> 00:15:17,709 and activists working on various subjects 455 00:15:17,710 --> 00:15:20,799 and working with a bunch of activists on 456 00:15:20,800 --> 00:15:23,289 tracking deportations from the U.K. 457 00:15:23,290 --> 00:15:25,509 This is the system the British government 458 00:15:25,510 --> 00:15:26,739 uses to deport people, 459 00:15:28,180 --> 00:15:29,409 which is basically where they hire 460 00:15:29,410 --> 00:15:31,659 private planes because it got 461 00:15:31,660 --> 00:15:35,079 embarrassing to do it on on a general 462 00:15:35,080 --> 00:15:37,299 general usual aircraft. 463 00:15:37,300 --> 00:15:38,739 And in the middle of the night, they put 464 00:15:38,740 --> 00:15:40,869 people onto these planes and they fly 465 00:15:40,870 --> 00:15:42,009 them out of the country. And this is a 466 00:15:42,010 --> 00:15:44,109 very hard system to see because 467 00:15:44,110 --> 00:15:45,339 it happens in the middle of the night at 468 00:15:45,340 --> 00:15:46,899 private terminals, airports. 469 00:15:46,900 --> 00:15:49,299 It happens within closed courtrooms 470 00:15:49,300 --> 00:15:51,369 and within privatized 471 00:15:51,370 --> 00:15:52,869 detention centers. 472 00:15:52,870 --> 00:15:55,059 And so working with 473 00:15:55,060 --> 00:15:56,469 some friends of mine who are architects 474 00:15:56,470 --> 00:15:58,480 and architectural modelers, we used 475 00:16:00,520 --> 00:16:02,349 planning permission, went to the council. 476 00:16:02,350 --> 00:16:04,119 We got the files on these kind of places. 477 00:16:04,120 --> 00:16:05,019 We did these kind of classic 478 00:16:05,020 --> 00:16:07,629 investigative journalism practices 479 00:16:07,630 --> 00:16:09,999 of finding eyewitness accounts, 480 00:16:10,000 --> 00:16:12,159 finding council documents and 481 00:16:12,160 --> 00:16:13,869 visiting and sketching places where we 482 00:16:13,870 --> 00:16:14,799 could where we weren't allowed to 483 00:16:14,800 --> 00:16:17,019 photograph in order to visualize them. 484 00:16:17,020 --> 00:16:19,089 So this is a courtroom in the center of 485 00:16:19,090 --> 00:16:20,619 London. In the UK, it's illegal to 486 00:16:20,620 --> 00:16:22,629 photograph courtrooms. 487 00:16:22,630 --> 00:16:24,169 So there's no photographs exist of this 488 00:16:24,170 --> 00:16:25,089 space. 489 00:16:25,090 --> 00:16:26,799 But there are very, very shiny 490 00:16:26,800 --> 00:16:28,599 architectural visualizations that we made 491 00:16:28,600 --> 00:16:30,579 of it. And this is an interesting 492 00:16:30,580 --> 00:16:32,109 courtroom because it's a courtroom in 493 00:16:32,110 --> 00:16:34,929 which special laws apply. 494 00:16:34,930 --> 00:16:36,669 The UK passed a special law for 495 00:16:36,670 --> 00:16:39,219 immigrants appeals commissions 496 00:16:39,220 --> 00:16:40,869 where there's suspicion of terrorist 497 00:16:40,870 --> 00:16:43,539 activity or some kind of security risk, 498 00:16:43,540 --> 00:16:45,549 where they can present secret evidence, 499 00:16:45,550 --> 00:16:47,619 which is evidence that the defendant 500 00:16:47,620 --> 00:16:48,999 and their legal team are not allowed to 501 00:16:49,000 --> 00:16:50,499 know about. It makes a complete mockery 502 00:16:50,500 --> 00:16:52,029 of the idea of a fair trial. 503 00:16:52,030 --> 00:16:53,799 And what's interesting is the way that 504 00:16:53,800 --> 00:16:55,239 the way in which that appears as 505 00:16:55,240 --> 00:16:57,309 architecture in spaces like this, because 506 00:16:57,310 --> 00:16:58,599 you can actually point to it. 507 00:16:58,600 --> 00:17:00,699 The divider on the left here is 508 00:17:00,700 --> 00:17:02,409 what separates the public viewing gallery 509 00:17:02,410 --> 00:17:04,029 from the spooks viewing gallery. 510 00:17:04,030 --> 00:17:06,279 And the closed witness box over there 511 00:17:06,280 --> 00:17:07,389 is so that members of the security 512 00:17:07,390 --> 00:17:09,489 personnel can Schizo's can 513 00:17:09,490 --> 00:17:10,479 appear without giving away their 514 00:17:10,480 --> 00:17:11,480 identities. 515 00:17:12,430 --> 00:17:14,650 And this culminated in a 516 00:17:16,180 --> 00:17:17,858 eight minute film you can watch online 517 00:17:17,859 --> 00:17:19,689 called Seamless Transitions, which 518 00:17:19,690 --> 00:17:20,769 documents these 519 00:17:22,480 --> 00:17:24,399 these various spaces. 520 00:17:24,400 --> 00:17:26,019 The courtroom, which you just saw, a 521 00:17:26,020 --> 00:17:27,969 detention center near Heathrow Airport 522 00:17:27,970 --> 00:17:30,219 and the final airport to which people 523 00:17:30,220 --> 00:17:32,019 are deported, which is Stansted, if 524 00:17:32,020 --> 00:17:34,059 anyone goes to the UK on holiday. 525 00:17:34,060 --> 00:17:36,159 And and 526 00:17:36,160 --> 00:17:37,989 what what again, I was trying to do in 527 00:17:37,990 --> 00:17:40,299 this these this film 528 00:17:40,300 --> 00:17:42,729 was not necessarily 529 00:17:42,730 --> 00:17:45,819 paint like just 530 00:17:45,820 --> 00:17:47,199 what you would traditionally do with 531 00:17:47,200 --> 00:17:48,099 photojournalism. 532 00:17:48,100 --> 00:17:49,089 So this was commissioned by the 533 00:17:49,090 --> 00:17:50,769 photographers gallery in London. 534 00:17:50,770 --> 00:17:53,019 And and it was 535 00:17:53,020 --> 00:17:54,849 when it first went on show that it was 536 00:17:54,850 --> 00:17:57,189 alongside a huge, 537 00:17:57,190 --> 00:17:59,319 amazing deep dives to the archive 538 00:17:59,320 --> 00:18:00,459 of something called the Black Star 539 00:18:00,460 --> 00:18:01,869 Archive, which is an archive of human 540 00:18:01,870 --> 00:18:03,839 rights photography from the 20th century. 541 00:18:03,840 --> 00:18:05,859 It's extraordinary, but it's very much 542 00:18:05,860 --> 00:18:07,869 what you think of when you think of 20th 543 00:18:07,870 --> 00:18:10,089 century photojournalism, which is the 544 00:18:10,090 --> 00:18:12,549 kind of sharp black and white photo 545 00:18:12,550 --> 00:18:14,859 a face in a crowd, this frozen moment, 546 00:18:14,860 --> 00:18:16,509 this moment from which we can supposedly 547 00:18:16,510 --> 00:18:18,819 draw like this particular story, 548 00:18:18,820 --> 00:18:21,249 this one signal thing which will explain 549 00:18:21,250 --> 00:18:22,250 everything else. 550 00:18:23,170 --> 00:18:25,239 And it's and it's a fantastic way of 551 00:18:25,240 --> 00:18:27,099 working. But in this, I kind of realized 552 00:18:27,100 --> 00:18:28,869 I was doing the opposite, which I was 553 00:18:28,870 --> 00:18:30,879 using the architecture of these spaces 554 00:18:30,880 --> 00:18:32,979 and also the way in which work 555 00:18:32,980 --> 00:18:34,659 like this is put together, which is 556 00:18:34,660 --> 00:18:37,029 usually based on research assemblage, 557 00:18:37,030 --> 00:18:39,339 co-operating with other artists 558 00:18:39,340 --> 00:18:41,529 visualizes and also cooperating 559 00:18:41,530 --> 00:18:43,659 with the software and architectural 560 00:18:43,660 --> 00:18:45,189 software that makes this is the same 561 00:18:45,190 --> 00:18:47,379 software that generates a huge amount 562 00:18:47,380 --> 00:18:49,149 of the built environment around us. 563 00:18:49,150 --> 00:18:51,279 And yet we still kind 564 00:18:51,280 --> 00:18:53,619 of assume there's one master architect 565 00:18:53,620 --> 00:18:54,939 at the heart of one of these things, 566 00:18:54,940 --> 00:18:56,709 rather than several thousand people 567 00:18:56,710 --> 00:18:59,169 working on Code AutoCAD or whoever it is 568 00:18:59,170 --> 00:19:00,999 who now have a huge stake in our built 569 00:19:01,000 --> 00:19:02,739 environment. So all of these different 570 00:19:02,740 --> 00:19:04,329 pressures come into play on trying to 571 00:19:04,330 --> 00:19:06,579 depict not a particular place, 572 00:19:06,580 --> 00:19:08,739 but the entire legal and 573 00:19:08,740 --> 00:19:10,569 infrastructural system that goes into 574 00:19:10,570 --> 00:19:12,729 creating the situation that 575 00:19:12,730 --> 00:19:15,159 results, which is horrific deportation 576 00:19:15,160 --> 00:19:16,160 system. 577 00:19:16,900 --> 00:19:19,209 So that's a very brief 578 00:19:19,210 --> 00:19:21,309 and swift run-Through of some 579 00:19:21,310 --> 00:19:22,420 of the sort of things I do. 580 00:19:24,340 --> 00:19:26,109 I want to talk a little bit more about 581 00:19:27,190 --> 00:19:29,530 about this project, the cloud index, 582 00:19:31,300 --> 00:19:32,829 which requires a little bit of history, 583 00:19:32,830 --> 00:19:33,939 which I hope will be interesting. 584 00:19:35,260 --> 00:19:37,509 This is this is one of my favorite 585 00:19:37,510 --> 00:19:38,979 people, Lewis Richardson 586 00:19:40,720 --> 00:19:43,539 in nineteen twelve thirteen. 587 00:19:43,540 --> 00:19:45,849 Richardson was a meteorologist 588 00:19:45,850 --> 00:19:48,219 who was working the media observatory 589 00:19:48,220 --> 00:19:50,589 in Scotland, doing a very, very early 590 00:19:50,590 --> 00:19:52,449 kind of weather calculation, mostly just 591 00:19:52,450 --> 00:19:54,249 recording the weather and trying to like 592 00:19:54,250 --> 00:19:56,319 think about what it might mean to start 593 00:19:56,320 --> 00:19:57,619 to have this kind of huge. 594 00:19:57,620 --> 00:19:59,719 Body of weather data, and 595 00:19:59,720 --> 00:20:01,459 then in 1914, he was called up for the 596 00:20:01,460 --> 00:20:03,889 First World War and he was a Quaker, 597 00:20:03,890 --> 00:20:04,879 so he was a pacifist. 598 00:20:04,880 --> 00:20:07,159 So he joined the ambulance division and 599 00:20:07,160 --> 00:20:09,259 spent the war running 600 00:20:09,260 --> 00:20:10,939 around with stretchers. 601 00:20:10,940 --> 00:20:12,709 The other thing he spent the war doing 602 00:20:12,710 --> 00:20:14,719 was that he'd taken a bunch of data, a 603 00:20:14,720 --> 00:20:16,519 bunch of readings from the observatory 604 00:20:16,520 --> 00:20:17,929 and from other observatories around 605 00:20:17,930 --> 00:20:19,639 Europe that had been gathering for the 606 00:20:19,640 --> 00:20:21,889 war. And over the course of two 607 00:20:21,890 --> 00:20:24,139 to three years during the war, 608 00:20:24,140 --> 00:20:26,359 while in trenches while under fire, he 609 00:20:26,360 --> 00:20:28,249 performed the very first mathematical 610 00:20:28,250 --> 00:20:31,189 calculation of a 24 hour weather forecast 611 00:20:31,190 --> 00:20:33,319 just over the area of northern Europe, 612 00:20:33,320 --> 00:20:35,389 just over the area that you see here. 613 00:20:35,390 --> 00:20:37,459 But he figured out what the matter was 614 00:20:37,460 --> 00:20:39,169 to take a bunch of historical weather 615 00:20:39,170 --> 00:20:41,389 data and advance that 616 00:20:41,390 --> 00:20:43,479 24 hours to to 617 00:20:43,480 --> 00:20:45,709 push that prediction forward into 618 00:20:45,710 --> 00:20:46,819 the future, which is an incredible 619 00:20:46,820 --> 00:20:48,199 achievement. Let's say it took him 620 00:20:48,200 --> 00:20:49,999 several years to do this with pencils and 621 00:20:50,000 --> 00:20:51,079 paper. 622 00:20:51,080 --> 00:20:53,179 And when he analyzed the 623 00:20:53,180 --> 00:20:54,679 results, because obviously the day he was 624 00:20:54,680 --> 00:20:57,079 doing was by that stage, almost a decade 625 00:20:57,080 --> 00:20:58,909 in the past, he turned out to be broadly 626 00:20:58,910 --> 00:21:00,139 correct. 627 00:21:00,140 --> 00:21:01,709 Some things were a bit more exaggerated, 628 00:21:01,710 --> 00:21:04,489 whatever. But but his his kind of 629 00:21:04,490 --> 00:21:06,199 his thesis worked. 630 00:21:06,200 --> 00:21:07,879 And because, of course, this was a time 631 00:21:07,880 --> 00:21:10,040 before what we now think of as computers. 632 00:21:11,630 --> 00:21:13,099 He didn't he thought this was a kind of 633 00:21:13,100 --> 00:21:15,229 achievement, but thought realizing its 634 00:21:15,230 --> 00:21:17,539 full potential was a very long way off. 635 00:21:17,540 --> 00:21:18,859 And he 636 00:21:20,200 --> 00:21:22,399 in the in this book, the one I showed 637 00:21:22,400 --> 00:21:24,649 before, he goes on this kind of beautiful 638 00:21:24,650 --> 00:21:26,929 little fantasy where he describes 639 00:21:26,930 --> 00:21:28,639 what it would take for his weather 640 00:21:28,640 --> 00:21:30,829 forecasting to actually 641 00:21:30,830 --> 00:21:32,779 achieve some kind of industrial scale and 642 00:21:32,780 --> 00:21:34,519 be economically useful. 643 00:21:34,520 --> 00:21:37,039 This is a painting of that fantasy 644 00:21:37,040 --> 00:21:39,319 by Stephen Conolan from 1986, 645 00:21:39,320 --> 00:21:40,320 where he 646 00:21:42,890 --> 00:21:45,469 Richardson described this vast globe 647 00:21:45,470 --> 00:21:47,539 on the inside of which would be painted a 648 00:21:47,540 --> 00:21:49,759 huge grid covering the entire earth, 649 00:21:49,760 --> 00:21:51,379 which weather forecasters could kind of 650 00:21:51,380 --> 00:21:53,659 point to read data off, 651 00:21:53,660 --> 00:21:55,639 calculate it, have it replaced. 652 00:21:55,640 --> 00:21:56,749 You can see there'll be teams of 653 00:21:56,750 --> 00:21:58,789 mathematicians then still called 654 00:21:58,790 --> 00:22:00,169 computers work in the middle. 655 00:22:00,170 --> 00:22:01,369 And then it's kind of amazing 656 00:22:01,370 --> 00:22:03,949 infrastructure of information, telegraphs 657 00:22:03,950 --> 00:22:07,009 and all the kind of communications 658 00:22:07,010 --> 00:22:08,809 infrastructure, the age pouring 659 00:22:08,810 --> 00:22:10,489 information into this place and pouring 660 00:22:10,490 --> 00:22:11,490 it back out again. 661 00:22:12,560 --> 00:22:14,839 And so this was he knew it was possible, 662 00:22:14,840 --> 00:22:16,129 but he thought, you know, mankind would 663 00:22:16,130 --> 00:22:17,240 never actually get to the point 664 00:22:18,460 --> 00:22:19,519 that it was possible. 665 00:22:19,520 --> 00:22:21,859 And mankind did get to that point 666 00:22:21,860 --> 00:22:25,159 in in 1950, 667 00:22:25,160 --> 00:22:26,539 when the very first 668 00:22:28,580 --> 00:22:30,769 mechanical calculation of a 24 669 00:22:30,770 --> 00:22:32,719 hour weather forecast was performed, 670 00:22:33,950 --> 00:22:35,899 which was done on the ENIAC, which is a 671 00:22:35,900 --> 00:22:37,039 computer that some of you have probably 672 00:22:37,040 --> 00:22:39,259 heard of, which is my personal 673 00:22:39,260 --> 00:22:40,260 favorite computer. 674 00:22:41,570 --> 00:22:43,639 The first stood first 675 00:22:43,640 --> 00:22:45,829 mechanical stored program computer built 676 00:22:45,830 --> 00:22:47,479 in the 1940s, the University of 677 00:22:47,480 --> 00:22:48,480 Pennsylvania. 678 00:22:49,190 --> 00:22:51,469 And this is the process for doing 679 00:22:51,470 --> 00:22:52,579 it. 680 00:22:52,580 --> 00:22:53,809 And they put together a team of 681 00:22:53,810 --> 00:22:56,149 meteorologists in 1948 682 00:22:56,150 --> 00:22:58,249 onwards. And in 1950, they ran 683 00:22:58,250 --> 00:23:00,349 this very, very first calculation or the 684 00:23:00,350 --> 00:23:02,389 very first successful calculation of a 685 00:23:02,390 --> 00:23:04,609 single day's weather forecasting over the 686 00:23:04,610 --> 00:23:05,630 northern United States. 687 00:23:06,770 --> 00:23:08,839 And when 688 00:23:08,840 --> 00:23:10,879 they did it, it took about three weeks to 689 00:23:10,880 --> 00:23:12,439 run. Right. 690 00:23:12,440 --> 00:23:14,029 Which doesn't sound so great. 691 00:23:14,030 --> 00:23:15,619 But you have to remember that the ENIAC 692 00:23:15,620 --> 00:23:17,929 broke down a lot of the time. 693 00:23:17,930 --> 00:23:20,149 And what they realized 694 00:23:20,150 --> 00:23:22,279 was that actually, if they subtracted all 695 00:23:22,280 --> 00:23:24,049 the time when the machine hadn't been 696 00:23:24,050 --> 00:23:25,999 running from the process, they got it 697 00:23:26,000 --> 00:23:28,249 down to 23 hours and 30 minutes. 698 00:23:28,250 --> 00:23:29,749 So they had very for the very first 699 00:23:29,750 --> 00:23:31,819 successfully time, calculated the 700 00:23:31,820 --> 00:23:33,679 weather faster than the weather itself 701 00:23:33,680 --> 00:23:34,859 was happening. 702 00:23:34,860 --> 00:23:37,189 Right. This is an incredibly 703 00:23:37,190 --> 00:23:38,389 significant moment. 704 00:23:38,390 --> 00:23:40,129 And John von Neumann, who worked on this 705 00:23:40,130 --> 00:23:42,589 project, wrote in the project log that 706 00:23:42,590 --> 00:23:44,509 we have achieved the dreams of Lewis 707 00:23:44,510 --> 00:23:46,639 Richardson. Computation now advances 708 00:23:46,640 --> 00:23:48,409 faster than the weather itself. 709 00:23:48,410 --> 00:23:50,029 And I like the fact that it happened on 710 00:23:50,030 --> 00:23:52,009 the ENIAC, which, as I say, is my 711 00:23:52,010 --> 00:23:53,689 favorite computer for those who aren't 712 00:23:53,690 --> 00:23:56,269 familiar with it. And here it is. 713 00:23:56,270 --> 00:23:59,269 And it was 714 00:23:59,270 --> 00:24:01,759 an early mainframe, so occupied 715 00:24:01,760 --> 00:24:03,649 two huge rooms, first at the University 716 00:24:03,650 --> 00:24:05,329 of Pennsylvania and later at the 717 00:24:06,620 --> 00:24:07,939 Plymouth Proving Grounds, Aberdeen 718 00:24:07,940 --> 00:24:10,039 Proving Grounds, where it 719 00:24:10,040 --> 00:24:11,689 was moved after the war. 720 00:24:11,690 --> 00:24:13,489 And it was you know, it required a lot of 721 00:24:13,490 --> 00:24:14,659 fiddling and plugging in. 722 00:24:14,660 --> 00:24:15,769 And if you read 723 00:24:16,910 --> 00:24:18,559 Turing's Cathedral, which is a great book 724 00:24:18,560 --> 00:24:20,839 about the kind of history of the early 725 00:24:20,840 --> 00:24:22,639 kind of computational programs, there's a 726 00:24:22,640 --> 00:24:24,199 beautiful quote in that from a guy called 727 00:24:24,200 --> 00:24:26,299 Harry Reid, who was an engineer 728 00:24:26,300 --> 00:24:28,369 who worked on this computer 729 00:24:28,370 --> 00:24:30,589 in which he says, I always 730 00:24:30,590 --> 00:24:32,209 thought of the ENIAC as a very personal 731 00:24:32,210 --> 00:24:34,129 computer. Now we think of a personal 732 00:24:34,130 --> 00:24:35,329 computer as something that we kind of 733 00:24:35,330 --> 00:24:36,559 carry around with us. 734 00:24:36,560 --> 00:24:37,999 But the ENIAC was a computer that you 735 00:24:38,000 --> 00:24:39,000 lived inside. 736 00:24:40,940 --> 00:24:42,679 It completely surrounded you. 737 00:24:42,680 --> 00:24:44,569 And actually for me like that, that 738 00:24:44,570 --> 00:24:46,129 division of personal computer is 739 00:24:46,130 --> 00:24:48,229 obviously and I think increasingly not 740 00:24:48,230 --> 00:24:50,419 true because we all live inside 741 00:24:50,420 --> 00:24:51,889 this kind of shell of computation. 742 00:24:51,890 --> 00:24:53,569 Now, we know that these are mostly 743 00:24:53,570 --> 00:24:55,969 terminals to other other 744 00:24:55,970 --> 00:24:57,549 connections, whoever owns them, whether. 745 00:24:57,550 --> 00:24:58,989 We build these things outwards or not, 746 00:24:58,990 --> 00:25:00,459 whether we're talking to satellites or 747 00:25:00,460 --> 00:25:02,499 not, the ENIAC has not contracted into 748 00:25:02,500 --> 00:25:04,599 these things, is actually expanded out 749 00:25:04,600 --> 00:25:06,550 into kind of vast shell around us. 750 00:25:07,750 --> 00:25:08,859 And the other thing that I really like 751 00:25:08,860 --> 00:25:10,359 about the NSA is, you know, it's one of 752 00:25:10,360 --> 00:25:12,759 these computers, this generation 753 00:25:12,760 --> 00:25:14,919 of computers were the last ones that 754 00:25:14,920 --> 00:25:17,589 were kind of truly physically legible. 755 00:25:17,590 --> 00:25:19,779 So computation now is legible to 756 00:25:19,780 --> 00:25:22,539 those of us, probably not me, who are 757 00:25:22,540 --> 00:25:24,489 really smart enough to kind of follow 758 00:25:24,490 --> 00:25:26,679 logically step by step, 759 00:25:26,680 --> 00:25:28,119 what kind of process is happening with 760 00:25:28,120 --> 00:25:29,589 the machines, with the ENIAC. 761 00:25:29,590 --> 00:25:30,999 You could do that physically with your 762 00:25:31,000 --> 00:25:33,279 eyes by looking at the blinking lights so 763 00:25:33,280 --> 00:25:34,989 that you could watch a computation kind 764 00:25:34,990 --> 00:25:38,349 of unfold on you across the walls. 765 00:25:38,350 --> 00:25:40,629 You could actually see that as it went 766 00:25:40,630 --> 00:25:42,189 from different parts of the machines to 767 00:25:42,190 --> 00:25:43,419 the to the architecture. 768 00:25:43,420 --> 00:25:44,679 You could kind of follow that. 769 00:25:44,680 --> 00:25:46,749 And that kind of legibility also seems to 770 00:25:46,750 --> 00:25:48,639 me something quite crucial to kind of go 771 00:25:48,640 --> 00:25:49,720 back to and reflect on. 772 00:25:51,370 --> 00:25:52,539 As I mentioned, one of the people who 773 00:25:52,540 --> 00:25:54,729 worked on that was John von Neumann, 774 00:25:54,730 --> 00:25:56,379 who I'm sure you're most familiar with, 775 00:25:56,380 --> 00:25:58,719 is the kind of inventor the von Neumann 776 00:25:58,720 --> 00:26:00,159 architecture worked on a lot of these 777 00:26:00,160 --> 00:26:01,959 early projects and obviously was working 778 00:26:01,960 --> 00:26:02,960 on both 779 00:26:04,060 --> 00:26:05,619 the weather project but also on the 780 00:26:05,620 --> 00:26:07,089 Manhattan Project, because this is here 781 00:26:07,090 --> 00:26:09,279 with with Oppenheimer, who 782 00:26:09,280 --> 00:26:11,409 ran the Manhattan Project and the 783 00:26:11,410 --> 00:26:13,209 same computer, the ENIAC was used for 784 00:26:13,210 --> 00:26:16,299 both the weather forecasting programs 785 00:26:16,300 --> 00:26:18,639 and for the development of nuclear 786 00:26:18,640 --> 00:26:21,189 bombs. These are the kind of two grand 787 00:26:21,190 --> 00:26:23,319 computational arms races that really 788 00:26:23,320 --> 00:26:25,479 kind of got computing going 789 00:26:25,480 --> 00:26:26,859 in the 1940s and 50s. 790 00:26:28,060 --> 00:26:30,549 And von Neumann had this line 791 00:26:30,550 --> 00:26:32,649 which which he did, which he wrote first, 792 00:26:32,650 --> 00:26:35,069 and a commentary on 793 00:26:35,070 --> 00:26:37,149 on the atomic on the weather project, 794 00:26:37,150 --> 00:26:38,889 but later recycled talk about the 795 00:26:38,890 --> 00:26:40,989 Manhattan Project, which was kind of his 796 00:26:40,990 --> 00:26:43,089 view of what computation could do 797 00:26:43,090 --> 00:26:44,379 and particularly computational 798 00:26:44,380 --> 00:26:46,299 prediction. And he said that all stable 799 00:26:46,300 --> 00:26:48,309 processes we should predict and or 800 00:26:48,310 --> 00:26:50,559 unstable processes we should control. 801 00:26:50,560 --> 00:26:51,879 And that was his view of what you could 802 00:26:51,880 --> 00:26:53,229 do with computation. 803 00:26:53,230 --> 00:26:54,759 You could you could model and control the 804 00:26:54,760 --> 00:26:56,529 world sufficient degree that you could 805 00:26:56,530 --> 00:26:57,579 predict the outcomes. 806 00:26:58,580 --> 00:26:59,580 So 807 00:27:01,310 --> 00:27:03,589 I wanted to to take some of those ideas 808 00:27:03,590 --> 00:27:05,149 of the weather and our computational 809 00:27:05,150 --> 00:27:06,829 predictive abilities and bring them 810 00:27:06,830 --> 00:27:09,019 somewhat up to date, and I was I got 811 00:27:09,020 --> 00:27:10,939 particularly fascinated with what's 812 00:27:10,940 --> 00:27:12,529 currently happening in neural networks 813 00:27:14,000 --> 00:27:16,629 because I studied statistics 814 00:27:16,630 --> 00:27:16,999 a while back. 815 00:27:17,000 --> 00:27:19,249 I studied II in 816 00:27:19,250 --> 00:27:21,769 the early noughties. 817 00:27:21,770 --> 00:27:23,419 But basically the idea of the 90s that 818 00:27:23,420 --> 00:27:25,489 was dying, so we were taught 819 00:27:25,490 --> 00:27:27,619 neural networks is a kind of historical 820 00:27:27,620 --> 00:27:29,419 anomaly, like a nice idea, some 821 00:27:29,420 --> 00:27:31,279 interesting theory, but basically a dead 822 00:27:31,280 --> 00:27:33,619 end. And the field kind of went quiet 823 00:27:33,620 --> 00:27:36,019 for 10 years and it's just starting to go 824 00:27:36,020 --> 00:27:37,129 kind of massive again. 825 00:27:37,130 --> 00:27:38,989 The last five years, neural networks have 826 00:27:38,990 --> 00:27:40,700 become and 827 00:27:41,870 --> 00:27:43,309 starting to kind of become inside 828 00:27:43,310 --> 00:27:44,959 everything essentially and rather 829 00:27:44,960 --> 00:27:46,579 interesting ways. And so I thought it was 830 00:27:46,580 --> 00:27:48,079 important to kind of reevaluate them and 831 00:27:48,080 --> 00:27:49,189 explore them again. 832 00:27:49,190 --> 00:27:51,349 And this is from a 833 00:27:51,350 --> 00:27:52,640 paper from 834 00:27:53,720 --> 00:27:55,489 the beginning of this year. 835 00:27:55,490 --> 00:27:56,779 Things move fast in this world 836 00:27:57,860 --> 00:28:00,169 of Facebook research paper 837 00:28:00,170 --> 00:28:02,719 about decisions, deep conversation, 838 00:28:02,720 --> 00:28:04,459 deep convolutional generational 839 00:28:04,460 --> 00:28:06,649 adversarial networks, which is something 840 00:28:06,650 --> 00:28:07,909 that they've been they've been pushing 841 00:28:07,910 --> 00:28:09,119 and using. 842 00:28:09,120 --> 00:28:11,839 And this is from a new network 843 00:28:11,840 --> 00:28:13,849 generation from that paper where they 844 00:28:13,850 --> 00:28:15,979 took thousands and 845 00:28:15,980 --> 00:28:18,409 thousands, thousands of bedrooms, uh, 846 00:28:18,410 --> 00:28:20,119 pictures of bedrooms. 847 00:28:20,120 --> 00:28:22,399 And they asked this network to simulate 848 00:28:22,400 --> 00:28:23,449 new bedrooms. 849 00:28:23,450 --> 00:28:25,579 So each of the images you see here is not 850 00:28:25,580 --> 00:28:27,619 a bedroom that exists in the world. 851 00:28:27,620 --> 00:28:29,869 It's a bedroom that has been dreamed up 852 00:28:29,870 --> 00:28:32,179 by a new network that's seen many, 853 00:28:32,180 --> 00:28:34,129 many pictures of bedrooms. 854 00:28:34,130 --> 00:28:36,229 And it's fascinating to 855 00:28:36,230 --> 00:28:37,490 see what it's like to try these things on 856 00:28:39,860 --> 00:28:42,019 and the more public 857 00:28:42,020 --> 00:28:44,189 image of neural networks, which 858 00:28:44,190 --> 00:28:45,649 is kind of everyone's a lot of people 859 00:28:45,650 --> 00:28:48,049 have seen is is the deep dream 860 00:28:48,050 --> 00:28:49,279 stuff from Google. 861 00:28:49,280 --> 00:28:51,469 And so what deep dream is, 862 00:28:51,470 --> 00:28:53,689 is an attempt to see back through 863 00:28:53,690 --> 00:28:55,939 these networks and try and understand how 864 00:28:55,940 --> 00:28:57,829 they think about and make images. 865 00:28:57,830 --> 00:28:59,959 So it's kind of instead of asking 866 00:28:59,960 --> 00:29:02,059 the the network 867 00:29:02,060 --> 00:29:03,649 to look at an image and tell you what it 868 00:29:03,650 --> 00:29:05,869 sees, it's asking it basically 869 00:29:05,870 --> 00:29:06,870 gives it 870 00:29:08,270 --> 00:29:10,429 blank or fixed images 871 00:29:10,430 --> 00:29:12,109 and it's kind of reverse them back 872 00:29:12,110 --> 00:29:14,209 through that network, kind of imagine and 873 00:29:14,210 --> 00:29:16,009 visualize what it can see in the world. 874 00:29:16,010 --> 00:29:17,149 And it turns out those things are 875 00:29:17,150 --> 00:29:18,419 terrifying. 876 00:29:18,420 --> 00:29:20,179 And I think I think this should be 877 00:29:20,180 --> 00:29:21,589 acknowledged more. 878 00:29:21,590 --> 00:29:23,849 And when 879 00:29:23,850 --> 00:29:25,669 the guys on the machine learning team at 880 00:29:25,670 --> 00:29:28,069 Google first came up with Deep Dream, 881 00:29:28,070 --> 00:29:29,929 apparently they showed it to Google and 882 00:29:29,930 --> 00:29:31,399 we're like, look what we made. 883 00:29:31,400 --> 00:29:33,579 And I was like, never release that. 884 00:29:35,420 --> 00:29:37,609 And in fact, they weren't allowed to. 885 00:29:37,610 --> 00:29:39,889 The dream leaked, first 886 00:29:39,890 --> 00:29:41,599 of all. And I don't know how, but it's 887 00:29:41,600 --> 00:29:42,569 interesting. 888 00:29:42,570 --> 00:29:44,059 And then they had to come out and say, 889 00:29:44,060 --> 00:29:45,889 OK, this is what this thing is. 890 00:29:45,890 --> 00:29:47,899 And everyone went, whoa, man, computer 891 00:29:47,900 --> 00:29:49,519 tripping balls and didn't 892 00:29:50,660 --> 00:29:51,949 and didn't. You know, there hasn't been a 893 00:29:51,950 --> 00:29:53,089 huge amount. 894 00:29:53,090 --> 00:29:55,559 And I don't think, 895 00:29:55,560 --> 00:29:57,679 you know, the people behind that 896 00:29:57,680 --> 00:29:58,939 are quite annoyed that there hasn't 897 00:29:58,940 --> 00:30:00,169 actually been a bit more thinking about 898 00:30:00,170 --> 00:30:01,699 what this kind of stuff entails, because 899 00:30:01,700 --> 00:30:02,819 it's an. 900 00:30:03,840 --> 00:30:06,029 What new networks are doing in the world 901 00:30:06,030 --> 00:30:08,549 is kind of interesting and 902 00:30:08,550 --> 00:30:10,289 you can do weird stuff with new networks 903 00:30:10,290 --> 00:30:12,419 and Visioneering actions and decisions, 904 00:30:12,420 --> 00:30:13,949 in particular this from the same paper. 905 00:30:13,950 --> 00:30:16,049 So you can do addition and subtraction. 906 00:30:16,050 --> 00:30:17,819 You can take a bunch of images associated 907 00:30:17,820 --> 00:30:19,949 with certain concepts, and you 908 00:30:19,950 --> 00:30:23,489 can then produce new images that combine 909 00:30:23,490 --> 00:30:24,959 elements of those things. 910 00:30:24,960 --> 00:30:26,489 So this is a network that has never seen 911 00:30:26,490 --> 00:30:28,799 a picture of a smiling man, but has seen 912 00:30:28,800 --> 00:30:30,989 the these other distinct 913 00:30:30,990 --> 00:30:31,829 stages. 914 00:30:31,830 --> 00:30:33,569 And you can do to check on that to 915 00:30:33,570 --> 00:30:35,789 produce new kinds of image outcomes. 916 00:30:35,790 --> 00:30:37,049 And so I thought, well, if a computer can 917 00:30:37,050 --> 00:30:38,249 dream of bedrooms and it can dream of 918 00:30:38,250 --> 00:30:40,139 smiling man, it can dream of all things 919 00:30:40,140 --> 00:30:41,369 and it can possibly also dream of the 920 00:30:41,370 --> 00:30:42,349 weather. 921 00:30:42,350 --> 00:30:43,890 And so I took. 922 00:30:47,190 --> 00:30:48,209 You might have noticed from the other 923 00:30:48,210 --> 00:30:49,379 work, some quite obsessed with satellite 924 00:30:49,380 --> 00:30:51,539 imagery, so I took eight years 925 00:30:51,540 --> 00:30:53,939 of satellite data 926 00:30:53,940 --> 00:30:55,109 of weather over the U.K. 927 00:30:55,110 --> 00:30:56,110 This is from the 928 00:30:57,300 --> 00:30:59,579 EU Space Programs Mattio site, 929 00:30:59,580 --> 00:31:01,799 which sits up there well above NUL 930 00:31:01,800 --> 00:31:04,139 Island, watching the Earth every day, 931 00:31:04,140 --> 00:31:06,779 several images a day of weather systems 932 00:31:06,780 --> 00:31:09,509 over the Western Hemisphere and 933 00:31:09,510 --> 00:31:12,659 chop that down into images over the UK, 934 00:31:12,660 --> 00:31:14,489 which came to about 16000 images, which I 935 00:31:14,490 --> 00:31:15,929 stress is really not enough for doing 936 00:31:15,930 --> 00:31:17,969 this kind of work. But I only had eight 937 00:31:17,970 --> 00:31:20,399 years of data on the other input, 938 00:31:20,400 --> 00:31:22,469 which was polling data from the UK 939 00:31:22,470 --> 00:31:24,119 on their voting intentions in the EU 940 00:31:24,120 --> 00:31:25,289 referendum. 941 00:31:25,290 --> 00:31:26,979 And I'm sorry about Brexit. 942 00:31:26,980 --> 00:31:28,419 I'm not going to speak of that again. 943 00:31:28,420 --> 00:31:29,639 I said I am because this is what this 944 00:31:29,640 --> 00:31:30,809 project is about. 945 00:31:30,810 --> 00:31:33,089 And this is this is a short 946 00:31:33,090 --> 00:31:34,589 snapshot. I actually had eight years of 947 00:31:34,590 --> 00:31:36,419 polling data. It's a bit sparse at the 948 00:31:36,420 --> 00:31:38,609 end, but it was enough to extrapolate 949 00:31:38,610 --> 00:31:40,319 eight years of information. 950 00:31:40,320 --> 00:31:42,479 And so I built one of these 951 00:31:42,480 --> 00:31:44,039 networks with the assistance of Jean 952 00:31:44,040 --> 00:31:46,029 Kogan, who might be around here somewhere 953 00:31:46,030 --> 00:31:49,199 and who's brilliant and 954 00:31:49,200 --> 00:31:52,049 fed. It is of 955 00:31:52,050 --> 00:31:54,209 whether images from space, eight 956 00:31:54,210 --> 00:31:56,489 years of images of cloud formations and 957 00:31:56,490 --> 00:31:58,619 eight years of this 958 00:31:58,620 --> 00:32:00,459 polling data about the EU referendum and 959 00:32:00,460 --> 00:32:02,099 my new network dreams new weather 960 00:32:02,100 --> 00:32:04,229 formations dream new kind of 961 00:32:04,230 --> 00:32:05,489 cloud patterns. 962 00:32:05,490 --> 00:32:06,899 And once this started to make these kind 963 00:32:06,900 --> 00:32:08,999 of associations, it was 964 00:32:09,000 --> 00:32:11,369 then possible to say, well, you've seen 965 00:32:11,370 --> 00:32:13,259 what the weather looks like on these 966 00:32:13,260 --> 00:32:15,269 certain occasions when the voting is 967 00:32:15,270 --> 00:32:17,229 trending in these particular directions. 968 00:32:17,230 --> 00:32:18,959 And so just as we did the addition with 969 00:32:18,960 --> 00:32:20,999 the smiling man cetera, let's do some 970 00:32:21,000 --> 00:32:23,159 additions on the voting and say, well, 971 00:32:23,160 --> 00:32:25,649 what should the weather look like if 972 00:32:25,650 --> 00:32:27,479 the entire country was to vote in one 973 00:32:27,480 --> 00:32:29,129 particular direction or in the other 974 00:32:29,130 --> 00:32:31,319 direction to generate weather 975 00:32:32,340 --> 00:32:34,469 potential wethers that we 976 00:32:34,470 --> 00:32:36,629 could simulate if we wish 977 00:32:36,630 --> 00:32:38,849 the democratic outcomes to be different 978 00:32:38,850 --> 00:32:40,919 and and the ones at 979 00:32:40,920 --> 00:32:42,959 the extremes are what they are. 980 00:32:42,960 --> 00:32:45,239 I actually kind of enjoyed more 981 00:32:45,240 --> 00:32:47,369 the outcomes that played around with 982 00:32:47,370 --> 00:32:48,809 undecided voters. 983 00:32:48,810 --> 00:32:50,039 This seemed to me actually something that 984 00:32:50,040 --> 00:32:52,979 really required more attention 985 00:32:52,980 --> 00:32:55,679 in the aftermath of 986 00:32:55,680 --> 00:32:57,659 the very divisive elections of the kind 987 00:32:57,660 --> 00:32:58,839 of various kinds. 988 00:32:58,840 --> 00:33:00,509 We've seen this year to try and put a 989 00:33:00,510 --> 00:33:03,359 little bit more emphasis on not on 990 00:33:03,360 --> 00:33:05,429 hard Brexit or 991 00:33:05,430 --> 00:33:07,589 Trump or whatever it is, but on the huge 992 00:33:07,590 --> 00:33:09,719 space within that of of indecision 993 00:33:09,720 --> 00:33:11,879 and unknowingness that actually guides 994 00:33:11,880 --> 00:33:13,979 and hugely influences a lot 995 00:33:13,980 --> 00:33:14,980 of our elections. 996 00:33:16,200 --> 00:33:17,849 And the reason that I thought this would 997 00:33:17,850 --> 00:33:20,399 be a worthwhile thing to do is because 998 00:33:20,400 --> 00:33:22,679 we can change the weather now 999 00:33:22,680 --> 00:33:24,389 if we if we want to. 1000 00:33:24,390 --> 00:33:26,819 This is Vincent Chafer in his laboratory 1001 00:33:26,820 --> 00:33:28,949 at General Electric in 1002 00:33:28,950 --> 00:33:30,179 the mid 1950s. 1003 00:33:30,180 --> 00:33:31,649 Vincent Schaefers, the man who invented 1004 00:33:31,650 --> 00:33:33,809 cloud seeding, which is the 1005 00:33:33,810 --> 00:33:36,269 realization that you can place into 1006 00:33:37,380 --> 00:33:39,719 into early form in clouds, crystals 1007 00:33:39,720 --> 00:33:42,329 of silver iodide that will cause those 1008 00:33:42,330 --> 00:33:44,219 clouds to precipitate. 1009 00:33:44,220 --> 00:33:46,409 You can you can you can raise 1010 00:33:46,410 --> 00:33:48,389 smoke levels. You can burn off fog. 1011 00:33:48,390 --> 00:33:50,609 You can enact quite large scale changes 1012 00:33:50,610 --> 00:33:51,899 to the atmosphere. 1013 00:33:51,900 --> 00:33:54,719 When Schaefers work started to appear 1014 00:33:54,720 --> 00:33:57,089 in the 1950s, there was a huge public 1015 00:33:57,090 --> 00:33:59,159 debate in the US and elsewhere 1016 00:33:59,160 --> 00:34:01,649 about what the possibilities 1017 00:34:01,650 --> 00:34:03,839 of weather modification might be. 1018 00:34:03,840 --> 00:34:05,519 There was a huge belief there's a huge 1019 00:34:05,520 --> 00:34:08,069 military belief that the weather would be 1020 00:34:08,070 --> 00:34:09,388 the weapon of the future. 1021 00:34:09,389 --> 00:34:11,729 And the US used weather modification 1022 00:34:11,730 --> 00:34:13,859 extensively in Vietnam and had a two 1023 00:34:13,860 --> 00:34:15,988 year program of seeding 1024 00:34:15,989 --> 00:34:18,059 with silver iodide clouds over 1025 00:34:18,060 --> 00:34:18,959 Vietnam. 1026 00:34:18,960 --> 00:34:21,059 And in order to basically obliterate 1027 00:34:21,060 --> 00:34:22,709 the Ho Chi Minh trail, they reckon they 1028 00:34:22,710 --> 00:34:24,809 extended the monsoon soon period 1029 00:34:24,810 --> 00:34:26,999 by an average of six weeks each year, 1030 00:34:27,000 --> 00:34:29,039 making it rain more and more and more. 1031 00:34:29,040 --> 00:34:31,259 And when that leaked 1032 00:34:31,260 --> 00:34:32,260 again leaking, 1033 00:34:33,780 --> 00:34:34,979 they stopped it immediately. 1034 00:34:34,980 --> 00:34:37,229 And there's been little 1035 00:34:37,230 --> 00:34:39,448 willingness in the West to continue 1036 00:34:39,449 --> 00:34:40,859 cloud seeding experiments ever since 1037 00:34:40,860 --> 00:34:43,079 then, but it doesn't stop other 1038 00:34:43,080 --> 00:34:45,178 people. These are Chinese, whether 1039 00:34:45,179 --> 00:34:46,678 making rockets. 1040 00:34:46,679 --> 00:34:49,229 The Chinese government has a 1041 00:34:49,230 --> 00:34:51,629 weather department of some 50000 1042 00:34:51,630 --> 00:34:53,369 employees whose job it is to make the 1043 00:34:53,370 --> 00:34:54,329 weather. 1044 00:34:54,330 --> 00:34:57,179 They active very much in agricultural 1045 00:34:57,180 --> 00:34:59,039 locations, either make it rain on the 1046 00:34:59,040 --> 00:35:01,109 crops or preventing it haling on crops. 1047 00:35:01,110 --> 00:35:03,239 But they also do a lot 1048 00:35:03,240 --> 00:35:06,239 of work clearing the air for 1049 00:35:06,240 --> 00:35:07,949 official parades, for the opening of the 1050 00:35:07,950 --> 00:35:09,959 Olympics, for these kinds of events. 1051 00:35:09,960 --> 00:35:10,960 So the 1052 00:35:12,090 --> 00:35:14,279 weather is is a tool that we could 1053 00:35:14,280 --> 00:35:16,439 be deploying, should we wish to, and 1054 00:35:16,440 --> 00:35:17,639 one that we might increasingly want to 1055 00:35:17,640 --> 00:35:18,640 think about deploying 1056 00:35:19,920 --> 00:35:22,169 in the in the not too distant future 1057 00:35:22,170 --> 00:35:23,339 when we have to ask ourselves quite 1058 00:35:23,340 --> 00:35:24,809 serious questions about what we want to 1059 00:35:24,810 --> 00:35:26,519 do with the environment, how we're 1060 00:35:26,520 --> 00:35:28,949 willing to use the things that we know, 1061 00:35:28,950 --> 00:35:31,559 how we wish to use those things, whether 1062 00:35:31,560 --> 00:35:33,659 we have the strength of 1063 00:35:33,660 --> 00:35:35,909 our beliefs in our ability 1064 00:35:35,910 --> 00:35:38,339 to act in the world 1065 00:35:38,340 --> 00:35:40,439 and and what do we do 1066 00:35:40,440 --> 00:35:42,659 this? And and it is 1067 00:35:42,660 --> 00:35:44,759 because I'm no longer convinced, 1068 00:35:44,760 --> 00:35:45,909 actually, that it's. 1069 00:35:45,910 --> 00:35:47,989 All to understand the world 1070 00:35:47,990 --> 00:35:50,119 just to processing kind 1071 00:35:50,120 --> 00:35:52,489 of vast amounts of information 1072 00:35:52,490 --> 00:35:54,639 that I don't think this is any longer 1073 00:35:54,640 --> 00:35:56,809 a productive, useful way of dealing with 1074 00:35:56,810 --> 00:35:58,759 the world. And but I realize that I kind 1075 00:35:58,760 --> 00:36:00,380 of wanted to retain the cloud 1076 00:36:01,760 --> 00:36:03,919 not as a kind of bad metaphor of kind of 1077 00:36:05,090 --> 00:36:06,439 loss, the engineering or corporate 1078 00:36:06,440 --> 00:36:08,359 control, and not as that kind of 1079 00:36:08,360 --> 00:36:09,409 materialist look. 1080 00:36:09,410 --> 00:36:12,349 I found the data cable kind of 1081 00:36:12,350 --> 00:36:14,969 like materialization of the world. 1082 00:36:14,970 --> 00:36:17,119 But as a way of thinking, the world 1083 00:36:17,120 --> 00:36:19,249 is recognizing that the cloud is 1084 00:36:19,250 --> 00:36:21,439 cloudy, that what it reveals 1085 00:36:21,440 --> 00:36:23,149 to us is often actually a kind of 1086 00:36:23,150 --> 00:36:25,189 difference and differentiation around the 1087 00:36:25,190 --> 00:36:27,319 world. And that is not a purely 1088 00:36:27,320 --> 00:36:29,449 kind of fixed channel of information 1089 00:36:29,450 --> 00:36:31,759 that's there to tell us how the world is, 1090 00:36:31,760 --> 00:36:34,009 but asks us to question it and thinking, 1091 00:36:34,010 --> 00:36:35,599 think about it more. 1092 00:36:35,600 --> 00:36:38,059 And this is increasingly necessary 1093 00:36:38,060 --> 00:36:40,489 because this informational 1094 00:36:40,490 --> 00:36:41,839 view of the world, this view that by 1095 00:36:41,840 --> 00:36:43,429 gathering information, we kind of make 1096 00:36:43,430 --> 00:36:45,229 more knowledge and more sense is 1097 00:36:45,230 --> 00:36:47,149 increasingly at risk. 1098 00:36:47,150 --> 00:36:49,459 And I'm sure, again, Moore's 1099 00:36:49,460 --> 00:36:51,349 Law is familiar to most of people in this 1100 00:36:51,350 --> 00:36:52,350 room, the world, that 1101 00:36:54,110 --> 00:36:55,609 the number of transistors doubles every 1102 00:36:55,610 --> 00:36:57,019 couple of years and some space to get 1103 00:36:57,020 --> 00:36:58,399 more and more power, more and more processing 1104 00:36:58,400 --> 00:36:59,400 power. 1105 00:37:00,350 --> 00:37:02,329 This is its inverse. 1106 00:37:02,330 --> 00:37:03,859 This is something that has been named a 1107 00:37:03,860 --> 00:37:05,419 room slower by people who work in the 1108 00:37:05,420 --> 00:37:07,009 pharmacological sciences. 1109 00:37:07,010 --> 00:37:09,169 This is the graph of 1110 00:37:09,170 --> 00:37:11,389 money spent on medical research 1111 00:37:11,390 --> 00:37:12,889 against drug discovery. 1112 00:37:12,890 --> 00:37:15,079 And it's going the other way. 1113 00:37:15,080 --> 00:37:16,880 The more money that's put into this, 1114 00:37:18,410 --> 00:37:20,389 the less and less we're learning. 1115 00:37:20,390 --> 00:37:23,419 And those various proposals 1116 00:37:23,420 --> 00:37:25,479 that have been put forward to 1117 00:37:25,480 --> 00:37:27,559 reverse analyzes of this, which is a long 1118 00:37:27,560 --> 00:37:28,969 running and long standing problem in 1119 00:37:28,970 --> 00:37:31,069 pharmacology, and I'm one of the 1120 00:37:31,070 --> 00:37:33,199 most prevalent theories is is essentially 1121 00:37:33,200 --> 00:37:34,969 too much information. 1122 00:37:34,970 --> 00:37:36,319 If you know how drug discovery is done 1123 00:37:36,320 --> 00:37:38,419 now, it's not basically people in 1124 00:37:38,420 --> 00:37:40,549 white coats in the lab and it basically 1125 00:37:40,550 --> 00:37:41,509 looks like a server farm. 1126 00:37:41,510 --> 00:37:43,459 It looks like large machines, it looks 1127 00:37:43,460 --> 00:37:45,409 like robots. And it looks like very, very 1128 00:37:45,410 --> 00:37:47,299 large datasets of information being 1129 00:37:47,300 --> 00:37:49,399 automatically run through these 1130 00:37:49,400 --> 00:37:52,459 high throughput screening machines 1131 00:37:52,460 --> 00:37:54,289 in order to test new drug reactions 1132 00:37:54,290 --> 00:37:55,869 against one another. 1133 00:37:55,870 --> 00:37:58,039 And this this problem seems 1134 00:37:58,040 --> 00:38:01,429 to be, uh, uh, 1135 00:38:01,430 --> 00:38:03,709 continuing. And one of the ways 1136 00:38:03,710 --> 00:38:05,359 in which science labs are dealing with it 1137 00:38:05,360 --> 00:38:07,609 is increasingly putting people back into 1138 00:38:07,610 --> 00:38:09,859 the process, not as some kind of, like, 1139 00:38:09,860 --> 00:38:12,439 airy fairy, uh, 1140 00:38:12,440 --> 00:38:14,059 you know, computers, bad people, a 1141 00:38:14,060 --> 00:38:16,159 magical kind of way, but simply 1142 00:38:16,160 --> 00:38:18,469 that there seems to be some other way of 1143 00:38:18,470 --> 00:38:20,779 thinking, the world that doesn't rely on 1144 00:38:20,780 --> 00:38:22,849 vast, vast datasets, but on plotting 1145 00:38:22,850 --> 00:38:24,979 slightly distinct tanon machine equations 1146 00:38:24,980 --> 00:38:25,980 through them. 1147 00:38:26,480 --> 00:38:28,759 The example I like of that is, 1148 00:38:28,760 --> 00:38:31,459 is the story of advanced chess. 1149 00:38:31,460 --> 00:38:33,769 Um, so this is the famous 1150 00:38:33,770 --> 00:38:36,889 game in 1997 1151 00:38:36,890 --> 00:38:39,469 between Kasparov and Deep Blue when 1152 00:38:39,470 --> 00:38:41,719 IBM spent a decade building a computer 1153 00:38:41,720 --> 00:38:43,909 to beat this one poor guy at 1154 00:38:43,910 --> 00:38:45,110 the only game he'd ever loved. 1155 00:38:46,250 --> 00:38:47,250 And 1156 00:38:49,010 --> 00:38:50,809 it was, you know, is this was this insane 1157 00:38:50,810 --> 00:38:53,029 moment because we basically like our 1158 00:38:53,030 --> 00:38:54,979 idea of intelligence is always like the 1159 00:38:54,980 --> 00:38:56,389 one thing we're still holding out. 1160 00:38:56,390 --> 00:38:58,339 And you'll notice in any discussion of AI 1161 00:38:58,340 --> 00:39:00,139 that what constitutes intelligence has 1162 00:39:00,140 --> 00:39:02,209 kind of receded for decades as 1163 00:39:02,210 --> 00:39:03,229 the machines have got better and we've 1164 00:39:03,230 --> 00:39:04,679 got know not that best. 1165 00:39:04,680 --> 00:39:06,409 You can continually move the barriers, 1166 00:39:06,410 --> 00:39:08,359 but at this point, chess was the thing. 1167 00:39:08,360 --> 00:39:09,949 Right. And it was a really shocking 1168 00:39:09,950 --> 00:39:12,079 moment when possibly the 1169 00:39:12,080 --> 00:39:13,909 greatest chess player at a time of all 1170 00:39:13,910 --> 00:39:16,579 time was beaten by this machine. 1171 00:39:16,580 --> 00:39:19,039 Um, and, 1172 00:39:19,040 --> 00:39:21,109 you know, Deep Blue was, 1173 00:39:21,110 --> 00:39:22,609 uh, was a brute force machine. 1174 00:39:22,610 --> 00:39:24,289 That's what it did. It basically was 1175 00:39:24,290 --> 00:39:25,760 capable of doing such 1176 00:39:27,770 --> 00:39:29,569 vast search through the kind of 1177 00:39:29,570 --> 00:39:32,059 development process space of 1178 00:39:32,060 --> 00:39:33,709 the game has evolved that it could 1179 00:39:33,710 --> 00:39:35,809 outthink any human by 1180 00:39:35,810 --> 00:39:37,699 a pure kind of brute force approach of 1181 00:39:37,700 --> 00:39:39,829 thinking ahead. And that that ultimately 1182 00:39:39,830 --> 00:39:40,939 seems to be right. We finally built 1183 00:39:40,940 --> 00:39:43,879 enough processors to beat a human. 1184 00:39:43,880 --> 00:39:45,799 Um, and what was interesting is that 1185 00:39:45,800 --> 00:39:46,909 while lots of people kind of like, oh, 1186 00:39:46,910 --> 00:39:49,519 God, that it the machines take over. 1187 00:39:49,520 --> 00:39:51,889 Kasparov came back just one year later 1188 00:39:51,890 --> 00:39:54,319 with a thing he called advanced chess, 1189 00:39:54,320 --> 00:39:56,089 which is a new kind of chess, which was 1190 00:39:56,090 --> 00:39:58,219 not humans versus 1191 00:39:58,220 --> 00:40:00,499 machines, but humans and machines versus 1192 00:40:00,500 --> 00:40:02,689 humans and machines, because it turns 1193 00:40:02,690 --> 00:40:05,059 out that today and, 1194 00:40:05,060 --> 00:40:06,049 you know, and this has become very 1195 00:40:06,050 --> 00:40:07,280 successful and 1196 00:40:09,080 --> 00:40:11,359 a you know, not 1197 00:40:11,360 --> 00:40:13,729 even a supercomputer today, advanced 1198 00:40:13,730 --> 00:40:15,259 computer will wipe the floor with any 1199 00:40:15,260 --> 00:40:17,899 human alive, but a human using 1200 00:40:17,900 --> 00:40:20,119 quite a relatively weak computer assisted 1201 00:40:20,120 --> 00:40:22,009 by it will wipe the floor with even the 1202 00:40:22,010 --> 00:40:23,419 biggest supercomputer. 1203 00:40:23,420 --> 00:40:26,329 There's a some kind of interesting 1204 00:40:26,330 --> 00:40:28,549 complementary ness to the way in 1205 00:40:28,550 --> 00:40:30,259 which this thought happens. 1206 00:40:30,260 --> 00:40:32,119 That seems to me one model of how we can 1207 00:40:32,120 --> 00:40:33,979 think about our relationship technology, 1208 00:40:33,980 --> 00:40:36,379 not as this kind of, uh, 1209 00:40:36,380 --> 00:40:37,939 completely oppositional thing, but as a 1210 00:40:37,940 --> 00:40:39,559 form of cooperation and kind of thinking 1211 00:40:39,560 --> 00:40:40,909 through problems together. 1212 00:40:40,910 --> 00:40:43,009 Um, but 1213 00:40:43,010 --> 00:40:45,109 at the same time, there's something also 1214 00:40:45,110 --> 00:40:46,429 that that's. Coming out of that at the 1215 00:40:46,430 --> 00:40:48,559 same time, because this is where we next 1216 00:40:48,560 --> 00:40:50,629 put intelligence, right, this is 1217 00:40:50,630 --> 00:40:52,489 into this is the Avago game 1218 00:40:53,510 --> 00:40:55,579 from early this year when Google's deep 1219 00:40:55,580 --> 00:40:58,159 mind, again, just a massive spoilsport 1220 00:40:58,160 --> 00:41:00,419 machine played 1221 00:41:00,420 --> 00:41:02,509 Lisa, all one of the greatest 1222 00:41:02,510 --> 00:41:04,659 go players in the world. 1223 00:41:04,660 --> 00:41:06,829 And despite a heroic fourth game, 1224 00:41:06,830 --> 00:41:07,830 third game, fourth time, 1225 00:41:09,080 --> 00:41:11,299 deep mind one and so go is now 1226 00:41:11,300 --> 00:41:13,399 again one of these things that we 1227 00:41:13,400 --> 00:41:15,469 held up as being such an important bit of 1228 00:41:15,470 --> 00:41:17,569 intelligence and is 1229 00:41:17,570 --> 00:41:19,699 now kind of has again been taken by 1230 00:41:19,700 --> 00:41:20,149 the machines. 1231 00:41:20,150 --> 00:41:22,159 And while I really look forward to 1232 00:41:22,160 --> 00:41:23,989 advance go and all the kind of 1233 00:41:23,990 --> 00:41:25,489 interesting things that will come out of 1234 00:41:25,490 --> 00:41:27,139 new ways to play it, there's something 1235 00:41:27,140 --> 00:41:29,149 fundamentally different between deep mind 1236 00:41:29,150 --> 00:41:30,150 and 1237 00:41:31,610 --> 00:41:32,610 deep blue. 1238 00:41:33,260 --> 00:41:35,119 And that is we don't understand how deep 1239 00:41:35,120 --> 00:41:37,379 blue, how four go. 1240 00:41:37,380 --> 00:41:39,049 So it does what it does. 1241 00:41:39,050 --> 00:41:41,929 We don't understand how the new network 1242 00:41:41,930 --> 00:41:44,269 actually beat these old test theories 1243 00:41:44,270 --> 00:41:45,379 and those kind of ways of watching the 1244 00:41:45,380 --> 00:41:47,209 game and picking them apart. 1245 00:41:47,210 --> 00:41:48,949 But essentially, this was a machine that 1246 00:41:48,950 --> 00:41:50,899 was trained first by a team of humans and 1247 00:41:50,900 --> 00:41:53,119 then trained against instances of itself 1248 00:41:53,120 --> 00:41:54,529 and its learning ability. 1249 00:41:54,530 --> 00:41:55,849 Well, it played against other instances, 1250 00:41:55,850 --> 00:41:57,969 so it was kind of exponential and 1251 00:41:57,970 --> 00:41:59,809 and it thought processes because it's in 1252 00:42:01,620 --> 00:42:03,289 a neural network in the way it's in train 1253 00:42:03,290 --> 00:42:05,989 that your network is unintelligible to us 1254 00:42:05,990 --> 00:42:07,909 at some kind of deep fundamental level, 1255 00:42:07,910 --> 00:42:09,919 we can maybe produce the kind of weird, 1256 00:42:09,920 --> 00:42:12,619 deep dream mindstate images 1257 00:42:12,620 --> 00:42:14,659 for for go, but they won't tell us what 1258 00:42:14,660 --> 00:42:16,759 the the process is that 1259 00:42:16,760 --> 00:42:18,080 allows this this to happen. 1260 00:42:20,840 --> 00:42:22,849 Everyone's heard this is the example of 1261 00:42:22,850 --> 00:42:25,039 of the strangeness of neural networks. 1262 00:42:25,040 --> 00:42:26,689 Uh, I've checked this story and it 1263 00:42:26,690 --> 00:42:28,249 doesn't appear to be true, but it's the 1264 00:42:28,250 --> 00:42:30,529 classical story of how strange 1265 00:42:30,530 --> 00:42:32,689 neural networks are and why they 1266 00:42:32,690 --> 00:42:34,969 do things that we can't quite understand, 1267 00:42:34,970 --> 00:42:37,339 which is that the US Army 1268 00:42:37,340 --> 00:42:39,559 commissioned a 1269 00:42:39,560 --> 00:42:42,319 neural network to recognize tanks. 1270 00:42:42,320 --> 00:42:44,329 They have a team of researchers, a tank, 1271 00:42:44,330 --> 00:42:45,829 a bunch of soldiers, and they said go and 1272 00:42:45,830 --> 00:42:47,629 hide in the woods, take loads of photos, 1273 00:42:47,630 --> 00:42:49,369 feed them into a neural network. 1274 00:42:49,370 --> 00:42:51,829 And we want a new network that can see 1275 00:42:51,830 --> 00:42:53,719 tanks in the trees, that can see 1276 00:42:53,720 --> 00:42:55,339 camouflaged artillery and so on, so 1277 00:42:55,340 --> 00:42:56,569 forth. So they went out. 1278 00:42:56,570 --> 00:42:57,709 They took these photos, they were 1279 00:42:57,710 --> 00:42:58,129 backslap. 1280 00:42:58,130 --> 00:43:00,199 They built a 1281 00:43:00,200 --> 00:43:01,969 neural network. They run all the images. 1282 00:43:01,970 --> 00:43:04,519 And, uh, 1283 00:43:04,520 --> 00:43:06,589 and it worked perfectly on 1284 00:43:06,590 --> 00:43:07,999 this kind of X thousand images that they 1285 00:43:08,000 --> 00:43:10,129 had the the new and every time 1286 00:43:10,130 --> 00:43:11,719 classified. Yes. A tank in this image. 1287 00:43:11,720 --> 00:43:12,829 Now, there's not even when the tank was 1288 00:43:12,830 --> 00:43:14,299 like hidden deep in the trees and 1289 00:43:14,300 --> 00:43:16,069 invisible as they went back to the army 1290 00:43:16,070 --> 00:43:18,289 to get new network and like here, go with 1291 00:43:18,290 --> 00:43:20,599 know and the Army put its new pictures 1292 00:43:20,600 --> 00:43:22,309 in and it completely failed. 1293 00:43:22,310 --> 00:43:23,869 It didn't work at all. 1294 00:43:23,870 --> 00:43:25,519 And when they went back, because they 1295 00:43:25,520 --> 00:43:27,679 were working with a relatively small 1296 00:43:27,680 --> 00:43:30,019 dataset, they looked at the images 1297 00:43:30,020 --> 00:43:31,609 again and they realized that the tank had 1298 00:43:31,610 --> 00:43:33,199 only been there in the morning. 1299 00:43:33,200 --> 00:43:34,699 And so the images of the tank had been 1300 00:43:34,700 --> 00:43:36,079 shot in the morning and all the images 1301 00:43:36,080 --> 00:43:37,219 without tanks have been shot in the 1302 00:43:37,220 --> 00:43:39,079 afternoon. And the new network is really, 1303 00:43:39,080 --> 00:43:40,789 really good work out whether it was 1304 00:43:40,790 --> 00:43:42,919 morning or afternoon and but 1305 00:43:42,920 --> 00:43:44,629 it completely and utterly failed to see 1306 00:43:44,630 --> 00:43:45,920 tanks. And 1307 00:43:47,000 --> 00:43:49,399 and these are the kind of stringencies 1308 00:43:49,400 --> 00:43:50,839 that occur when using technologies where 1309 00:43:50,840 --> 00:43:53,089 we don't quite see the way the machine 1310 00:43:53,090 --> 00:43:55,040 sees the world and we're building 1311 00:43:56,450 --> 00:43:58,609 new networks into 1312 00:43:58,610 --> 00:43:59,719 all of our things. 1313 00:43:59,720 --> 00:44:02,449 Now, this is Nvidia's program 1314 00:44:02,450 --> 00:44:04,609 to, um, to train 1315 00:44:04,610 --> 00:44:06,229 cars, to drive or they basically just put 1316 00:44:06,230 --> 00:44:08,029 the neural network in the car connected 1317 00:44:08,030 --> 00:44:09,289 to the computer and driven around for a 1318 00:44:09,290 --> 00:44:10,789 while. So it learns to recognize the 1319 00:44:10,790 --> 00:44:12,409 world around it without any external 1320 00:44:12,410 --> 00:44:13,639 cues. It doesn't know the rules of the 1321 00:44:13,640 --> 00:44:15,589 world, rules of the road. 1322 00:44:15,590 --> 00:44:17,149 It doesn't know what anything else looks 1323 00:44:17,150 --> 00:44:19,009 like. It's just watched humans drive for 1324 00:44:19,010 --> 00:44:21,139 so long, it figures it out for itself. 1325 00:44:21,140 --> 00:44:23,749 Um, these things are becoming 1326 00:44:23,750 --> 00:44:25,969 inside the things around us all the time. 1327 00:44:25,970 --> 00:44:28,639 And and we're actively building 1328 00:44:28,640 --> 00:44:30,859 the tools of our own mystification. 1329 00:44:30,860 --> 00:44:32,359 This I find particularly strange. 1330 00:44:32,360 --> 00:44:34,459 This is from a Google brain experiment 1331 00:44:34,460 --> 00:44:36,169 just a couple of months old, where they 1332 00:44:36,170 --> 00:44:37,170 are trained 1333 00:44:38,240 --> 00:44:40,309 to neural networks to hide their 1334 00:44:40,310 --> 00:44:42,709 communications from a third party. 1335 00:44:42,710 --> 00:44:45,159 So they and Alice 1336 00:44:45,160 --> 00:44:47,509 and Bob in this arrangement evolve 1337 00:44:47,510 --> 00:44:49,579 their own cryptographic system 1338 00:44:49,580 --> 00:44:51,829 to protect their communications 1339 00:44:51,830 --> 00:44:53,959 from Eve. And I myself think it's 1340 00:44:53,960 --> 00:44:56,119 a really great idea that we should 1341 00:44:56,120 --> 00:44:57,709 train the machines to be able to talk in 1342 00:44:57,710 --> 00:44:59,419 secret behind our backs. 1343 00:44:59,420 --> 00:45:00,420 I think this is 1344 00:45:02,900 --> 00:45:04,999 struggling to see and 1345 00:45:05,000 --> 00:45:06,709 see why we would want to do this. 1346 00:45:06,710 --> 00:45:07,710 Um. 1347 00:45:09,100 --> 00:45:11,529 There's this idea, though, of course, of 1348 00:45:11,530 --> 00:45:12,909 being able to gather all the more data 1349 00:45:12,910 --> 00:45:13,929 about the world, make it 1350 00:45:15,970 --> 00:45:18,129 to collect everything in 1351 00:45:18,130 --> 00:45:19,539 order to make better decisions, is, of 1352 00:45:19,540 --> 00:45:22,419 course, one that we see across, 1353 00:45:22,420 --> 00:45:23,429 across and 1354 00:45:24,550 --> 00:45:25,899 across domains. 1355 00:45:25,900 --> 00:45:28,119 This is the this is the approach 1356 00:45:28,120 --> 00:45:29,349 of surveillance agencies. 1357 00:45:29,350 --> 00:45:31,329 This belief that by gathering all of the 1358 00:45:31,330 --> 00:45:33,099 information there will be able to make 1359 00:45:33,100 --> 00:45:34,479 better decisions. They will be able to 1360 00:45:34,480 --> 00:45:35,769 pull out and they will build a model of 1361 00:45:35,770 --> 00:45:37,569 the world that is so complete that 1362 00:45:37,570 --> 00:45:39,699 they'll have a full understanding of 1363 00:45:39,700 --> 00:45:41,350 it on which they will be able to act. 1364 00:45:42,880 --> 00:45:45,039 And it strikes me that 1365 00:45:45,040 --> 00:45:47,289 this this is 1366 00:45:47,290 --> 00:45:49,059 not actually all too different from the 1367 00:45:49,060 --> 00:45:52,209 way in which we also oppose this belief 1368 00:45:52,210 --> 00:45:54,789 that actually the belief in 1369 00:45:54,790 --> 00:45:57,189 information itself being somehow 1370 00:45:57,190 --> 00:45:59,919 sufficient to change opinions 1371 00:45:59,920 --> 00:46:00,920 to 1372 00:46:02,020 --> 00:46:04,869 to sort of enact some greater good, 1373 00:46:04,870 --> 00:46:06,939 naturally on its own behalf is also 1374 00:46:06,940 --> 00:46:09,429 the the approach of 1375 00:46:09,430 --> 00:46:11,829 transparency of 1376 00:46:11,830 --> 00:46:13,329 kind of full information release 1377 00:46:14,390 --> 00:46:16,090 that the the 1378 00:46:17,200 --> 00:46:19,299 NSA and WikiLeaks essentially believe 1379 00:46:19,300 --> 00:46:21,219 the same thing, which there is some 1380 00:46:21,220 --> 00:46:23,109 smoking gun information at the heart of 1381 00:46:23,110 --> 00:46:24,669 the world, that if we only bring it to 1382 00:46:24,670 --> 00:46:26,169 light, everything will kind of magically 1383 00:46:26,170 --> 00:46:27,279 be made better. 1384 00:46:27,280 --> 00:46:29,679 And that that that seems to me to be 1385 00:46:29,680 --> 00:46:31,419 an increasingly insufficient way of 1386 00:46:31,420 --> 00:46:32,529 viewing the world. 1387 00:46:32,530 --> 00:46:34,629 This is, as a small 1388 00:46:34,630 --> 00:46:36,159 example, Facebook's analysis of hot 1389 00:46:36,160 --> 00:46:38,439 topics for November 2016 with 1390 00:46:38,440 --> 00:46:40,569 WikiLeaks right up there in the old men 1391 00:46:40,570 --> 00:46:43,329 quadrant of most discussed topics 1392 00:46:43,330 --> 00:46:44,649 from the time of the election. 1393 00:46:44,650 --> 00:46:47,019 And I'm a big fan of Julian 1394 00:46:47,020 --> 00:46:49,269 Assange's original 1395 00:46:49,270 --> 00:46:51,609 network analysis of secrecy that posed 1396 00:46:51,610 --> 00:46:53,619 the kind of leaking as this kind of like 1397 00:46:53,620 --> 00:46:56,089 gripped in authoritarian regimes 1398 00:46:56,090 --> 00:46:57,099 that would kind of make them grind to a 1399 00:46:57,100 --> 00:46:59,109 halt. But it does appear to me at the 1400 00:46:59,110 --> 00:47:01,179 moment that right now it's kind 1401 00:47:01,180 --> 00:47:03,309 of merely oiling a machine that none 1402 00:47:03,310 --> 00:47:05,109 of us want to keep running, that we're 1403 00:47:05,110 --> 00:47:07,299 locked in a kind of deadlock 1404 00:47:07,300 --> 00:47:08,469 of an information 1405 00:47:10,000 --> 00:47:12,969 war, that we're both playing 1406 00:47:12,970 --> 00:47:15,099 against each other in ways that 1407 00:47:15,100 --> 00:47:16,869 are not going to break us out of this 1408 00:47:16,870 --> 00:47:18,670 this opacity, transparency, 1409 00:47:19,780 --> 00:47:21,460 dialectic any time soon. 1410 00:47:23,270 --> 00:47:25,439 And all around us, this 1411 00:47:25,440 --> 00:47:27,599 it seems to me that 1412 00:47:27,600 --> 00:47:30,179 it's becoming harder for these predictive 1413 00:47:30,180 --> 00:47:31,979 mathematical information BAE Systems to 1414 00:47:31,980 --> 00:47:34,079 work, and we need to ask very specific 1415 00:47:34,080 --> 00:47:36,299 questions about what that's doing 1416 00:47:36,300 --> 00:47:38,939 to us and to our societies. 1417 00:47:38,940 --> 00:47:40,230 There's a huge fuss around 1418 00:47:41,370 --> 00:47:42,370 polling 1419 00:47:43,650 --> 00:47:46,229 and the poor 1420 00:47:46,230 --> 00:47:47,699 performance of polls, both in the 1421 00:47:47,700 --> 00:47:50,040 referendum and Trump and everything else. 1422 00:47:51,570 --> 00:47:53,759 But, um, but most of that seem 1423 00:47:53,760 --> 00:47:54,869 to focus on the fact that, like, the 1424 00:47:54,870 --> 00:47:56,819 mathematical models were wrong and they 1425 00:47:56,820 --> 00:47:58,109 have to be revised and we kind of build 1426 00:47:58,110 --> 00:47:59,819 better ones rather than saying these 1427 00:47:59,820 --> 00:48:01,139 things are things that actually poison 1428 00:48:01,140 --> 00:48:03,539 the discourse as part of the process, 1429 00:48:03,540 --> 00:48:05,639 that when you are running polls, 1430 00:48:05,640 --> 00:48:07,469 running metrics, running faulty models 1431 00:48:07,470 --> 00:48:09,569 like this all of the time alongside 1432 00:48:09,570 --> 00:48:11,309 everything, you're actually shaping the 1433 00:48:11,310 --> 00:48:12,679 outcomes themselves. 1434 00:48:12,680 --> 00:48:14,819 So when polls are done during a campaign, 1435 00:48:14,820 --> 00:48:16,319 they don't just say what people are going 1436 00:48:16,320 --> 00:48:18,089 to do at the end. They change people's 1437 00:48:18,090 --> 00:48:20,099 behavior during the campaign. 1438 00:48:20,100 --> 00:48:22,499 And they also change what is politically 1439 00:48:22,500 --> 00:48:24,599 possible because people run polls and 1440 00:48:24,600 --> 00:48:25,739 they say, oh, people don't like that. 1441 00:48:25,740 --> 00:48:27,059 We're not even going to put this out into 1442 00:48:27,060 --> 00:48:28,439 the public discourse. 1443 00:48:28,440 --> 00:48:29,999 Modeling and predicting in this way 1444 00:48:30,000 --> 00:48:32,279 restricts the kind of actual 1445 00:48:32,280 --> 00:48:34,169 possibility of discourse around us. 1446 00:48:35,800 --> 00:48:38,049 The debate around fake news 1447 00:48:38,050 --> 00:48:40,269 is a kind of startlingly 1448 00:48:40,270 --> 00:48:42,459 interesting one, but one that is not 1449 00:48:42,460 --> 00:48:44,769 also historically new or novel or 1450 00:48:44,770 --> 00:48:46,419 should not be and also needs to be, I 1451 00:48:46,420 --> 00:48:48,039 think, very specifically, historically 1452 00:48:48,040 --> 00:48:50,409 situated as something that as an idea 1453 00:48:50,410 --> 00:48:52,479 of suddenly that we can't know 1454 00:48:52,480 --> 00:48:54,639 this stuff or that it doesn't emerge 1455 00:48:54,640 --> 00:48:56,919 from quite specific ways 1456 00:48:56,920 --> 00:48:58,329 of seeing the world. I was particularly 1457 00:48:58,330 --> 00:49:00,699 struck by 1458 00:49:00,700 --> 00:49:02,769 and some of the reporting in recent 1459 00:49:02,770 --> 00:49:04,960 weeks that went on and on about 1460 00:49:06,400 --> 00:49:07,989 the fact a lot of this fake news stuff 1461 00:49:07,990 --> 00:49:09,609 emerging from the former Yugoslav 1462 00:49:09,610 --> 00:49:10,780 Republic of Macedonia 1463 00:49:11,890 --> 00:49:13,419 that cited it, particularly in this 1464 00:49:13,420 --> 00:49:15,639 place, without any particular context 1465 00:49:15,640 --> 00:49:17,919 around the political situation 1466 00:49:17,920 --> 00:49:19,779 in that place, is possibly being related. 1467 00:49:19,780 --> 00:49:20,919 So the story, if you don't know it, is 1468 00:49:20,920 --> 00:49:22,659 that lots of kids in Macedonia and 1469 00:49:22,660 --> 00:49:24,939 particularly in this one tiny village set 1470 00:49:24,940 --> 00:49:26,859 up those these websites, which are 1471 00:49:26,860 --> 00:49:28,539 getting more hits on Facebook, The New 1472 00:49:28,540 --> 00:49:29,949 York Times or any kind of educational 1473 00:49:29,950 --> 00:49:32,649 media by completely making up stories 1474 00:49:32,650 --> 00:49:33,969 and are getting tons and tons of ad 1475 00:49:33,970 --> 00:49:35,049 revenue. 1476 00:49:35,050 --> 00:49:36,519 And it strikes me that this is a place in 1477 00:49:36,520 --> 00:49:38,979 which to have a government 1478 00:49:38,980 --> 00:49:40,839 in Macedonia, which for the last 10 years 1479 00:49:40,840 --> 00:49:42,969 has pursued a policy of fake 1480 00:49:42,970 --> 00:49:44,409 news as government policy. 1481 00:49:44,410 --> 00:49:47,529 This is part of Skopje 2014, 1482 00:49:47,530 --> 00:49:49,629 part of the Macedonian government 1483 00:49:49,630 --> 00:49:52,149 project to build an entirely 1484 00:49:52,150 --> 00:49:55,659 fake history for Macedonia 1485 00:49:55,660 --> 00:49:57,399 by erecting statues of Alexander the 1486 00:49:57,400 --> 00:49:59,469 Great and Philip of 1487 00:49:59,470 --> 00:50:01,689 Macedon and classicists buildings 1488 00:50:01,690 --> 00:50:03,189 with kind of Hellenic columns and this 1489 00:50:03,190 --> 00:50:05,289 kind of stuff in a completely kind 1490 00:50:05,290 --> 00:50:06,999 of a historical move that's been opposed 1491 00:50:07,000 --> 00:50:08,259 by lots of people in the country. 1492 00:50:08,260 --> 00:50:09,969 But as a result of a kind of huge, 1493 00:50:09,970 --> 00:50:12,009 resurgent nationalism and nationalism, 1494 00:50:12,010 --> 00:50:14,109 that actually depends on feeding a kind 1495 00:50:14,110 --> 00:50:15,699 of fake stories. 1496 00:50:15,700 --> 00:50:17,649 And we seem to be optimizing a lot of 1497 00:50:17,650 --> 00:50:19,149 other stuff for this as well. 1498 00:50:19,150 --> 00:50:21,399 And we seem to kind of desire this 1499 00:50:21,400 --> 00:50:24,309 inability to fully and 1500 00:50:24,310 --> 00:50:25,989 to see and understand the world. 1501 00:50:25,990 --> 00:50:28,159 These are images that 1502 00:50:28,160 --> 00:50:30,549 our researcher Robert Smith discovered 1503 00:50:30,550 --> 00:50:32,619 in his own photo stream when 1504 00:50:32,620 --> 00:50:34,719 he he posted originally 1505 00:50:34,720 --> 00:50:35,920 the two pictures on the left. 1506 00:50:36,970 --> 00:50:39,039 And Google's I think it's 1507 00:50:39,040 --> 00:50:40,569 called like automatical algorithm or 1508 00:50:40,570 --> 00:50:42,819 something, produced the image on 1509 00:50:42,820 --> 00:50:45,009 the right, which is an image that 1510 00:50:45,010 --> 00:50:46,539 combines them both to have both people 1511 00:50:46,540 --> 00:50:48,909 smiling. But it is an image of a moment 1512 00:50:48,910 --> 00:50:50,079 that didn't happen. 1513 00:50:50,080 --> 00:50:52,269 Right now, it's 1514 00:50:52,270 --> 00:50:54,429 a it's a memory of something that 1515 00:50:54,430 --> 00:50:56,139 never actually occurred. 1516 00:50:56,140 --> 00:50:58,359 And likewise, Adobe is 1517 00:50:58,360 --> 00:50:59,469 pushing a thing you might think would 1518 00:50:59,470 --> 00:51:01,629 Vercoe, which is a way 1519 00:51:01,630 --> 00:51:03,999 of editing 1520 00:51:04,000 --> 00:51:05,979 sound as you would edit text. 1521 00:51:05,980 --> 00:51:08,769 So a basically a tool for 1522 00:51:08,770 --> 00:51:09,819 a kind of public tool, a kind of 1523 00:51:09,820 --> 00:51:12,069 Photoshop for audio essentially 1524 00:51:12,070 --> 00:51:13,659 that allows anyone to kind of create 1525 00:51:13,660 --> 00:51:15,129 these kind of things. Our entire, 1526 00:51:16,300 --> 00:51:18,609 uh, a lot of our technological advances 1527 00:51:18,610 --> 00:51:20,649 seem to be in order to kind of create and 1528 00:51:20,650 --> 00:51:22,329 change and mold the fabric of reality in 1529 00:51:22,330 --> 00:51:24,069 ways that are not always visible and 1530 00:51:24,070 --> 00:51:26,319 transparent to show the visitor. 1531 00:51:26,320 --> 00:51:28,239 And the outcome of that is a world in 1532 00:51:28,240 --> 00:51:29,680 which people do not believe anything 1533 00:51:30,730 --> 00:51:33,969 when when complete authority of 1534 00:51:33,970 --> 00:51:36,129 any kind of trust in institutions. 1535 00:51:36,130 --> 00:51:37,119 No, I'm No. 1536 00:51:37,120 --> 00:51:38,739 One for blind trust in institutions 1537 00:51:38,740 --> 00:51:39,789 either. 1538 00:51:39,790 --> 00:51:41,619 But the fact that when in the UK you had 1539 00:51:41,620 --> 00:51:42,820 a referendum and 1540 00:51:43,840 --> 00:51:45,249 something like more than twenty five 1541 00:51:45,250 --> 00:51:47,679 percent of views of 1542 00:51:47,680 --> 00:51:49,479 participants believe that was going to be 1543 00:51:49,480 --> 00:51:51,759 rigged by the security services 1544 00:51:51,760 --> 00:51:53,769 is a very worrying situation to be in. 1545 00:51:54,800 --> 00:51:57,010 And that is a result of 1546 00:51:58,390 --> 00:52:00,969 a kind of a crisis of literacy, 1547 00:52:00,970 --> 00:52:02,679 an inability to deal with these 1548 00:52:02,680 --> 00:52:04,989 increasingly complex narratives 1549 00:52:04,990 --> 00:52:07,359 that are supplied by our technologies. 1550 00:52:07,360 --> 00:52:10,119 More information, as we saw in the 1551 00:52:10,120 --> 00:52:11,649 pharmacological research we've seen, the 1552 00:52:11,650 --> 00:52:13,749 neural networks doesn't necessarily lead 1553 00:52:13,750 --> 00:52:15,429 to better decisions and in fact, often 1554 00:52:15,430 --> 00:52:18,059 kind of hardens people against them. 1555 00:52:18,060 --> 00:52:20,419 The sociologists, Fredric Jameson 1556 00:52:20,420 --> 00:52:22,599 in his work Cognitive Mapping describe 1557 00:52:22,600 --> 00:52:24,729 this is the scope conspiracy 1558 00:52:24,730 --> 00:52:26,679 theory in particular as this attempt 1559 00:52:26,680 --> 00:52:28,839 cognitively to map 1560 00:52:28,840 --> 00:52:30,439 landscapes of information which are 1561 00:52:30,440 --> 00:52:32,739 simply too vast for people to grasp 1562 00:52:32,740 --> 00:52:34,899 that no one person can hold this 1563 00:52:34,900 --> 00:52:37,209 entire system of information inside 1564 00:52:37,210 --> 00:52:38,709 their head. It's the same problem as big 1565 00:52:38,710 --> 00:52:40,239 data. It's also the same problem we 1566 00:52:40,240 --> 00:52:42,309 encounter every day on the 1567 00:52:42,310 --> 00:52:44,499 Internet because we are continually 1568 00:52:44,500 --> 00:52:46,749 faced with stuff that doesn't map into 1569 00:52:46,750 --> 00:52:47,829 our model of the world. 1570 00:52:47,830 --> 00:52:49,479 And that is existentially terrifying 1571 00:52:49,480 --> 00:52:51,459 because we're quite weak, you know, 1572 00:52:51,460 --> 00:52:53,709 lizard brain creatures, and 1573 00:52:53,710 --> 00:52:55,419 that we have not equipped ourselves with 1574 00:52:55,420 --> 00:52:57,549 literacy for dealing with complicated 1575 00:52:57,550 --> 00:52:59,859 and often paradoxical information. 1576 00:52:59,860 --> 00:53:01,449 And as a child of the kind of hippie 1577 00:53:01,450 --> 00:53:04,149 Internet. Right, this is deeply worrying 1578 00:53:04,150 --> 00:53:05,379 because for me, the Internet has always 1579 00:53:05,380 --> 00:53:08,169 been a place of incredible emancipatory 1580 00:53:08,170 --> 00:53:10,749 possibility, a place in which 1581 00:53:10,750 --> 00:53:12,309 you have access to all of this 1582 00:53:12,310 --> 00:53:14,439 information and it makes the 1583 00:53:14,440 --> 00:53:16,149 world more comprehensible to you. 1584 00:53:16,150 --> 00:53:18,789 It gives you greater agency, 1585 00:53:18,790 --> 00:53:21,099 and that allows you to make 1586 00:53:21,100 --> 00:53:22,869 more social better or whatever. 1587 00:53:22,870 --> 00:53:24,279 You seem very naive. 1588 00:53:24,280 --> 00:53:26,439 You, it turns out, but 1589 00:53:26,440 --> 00:53:28,719 a really important one to get across 1590 00:53:28,720 --> 00:53:29,979 because we're living in a world where 1591 00:53:29,980 --> 00:53:31,239 information has been increasingly 1592 00:53:31,240 --> 00:53:33,369 available for a long time 1593 00:53:33,370 --> 00:53:34,419 now and is in. 1594 00:53:34,420 --> 00:53:36,549 Recently riven by fundamentalisms and 1595 00:53:36,550 --> 00:53:38,409 deeply entrenched differing points of 1596 00:53:38,410 --> 00:53:41,079 view, so we need to think very carefully 1597 00:53:41,080 --> 00:53:43,239 about what our 1598 00:53:43,240 --> 00:53:45,339 what our responses 1599 00:53:45,340 --> 00:53:47,469 are, what are the new literacies that we 1600 00:53:47,470 --> 00:53:49,669 need to build, not to understand 1601 00:53:49,670 --> 00:53:51,999 raw information, but to understand these 1602 00:53:52,000 --> 00:53:54,579 these differences between information. 1603 00:53:54,580 --> 00:53:56,769 And there was a I 1604 00:53:56,770 --> 00:53:58,449 took the title of this talk from a piece 1605 00:53:58,450 --> 00:54:00,519 in The New York Times from 1606 00:54:00,520 --> 00:54:02,589 a few weeks ago by the director of 1607 00:54:02,590 --> 00:54:04,779 the Global Weather Forecast, Global 1608 00:54:04,780 --> 00:54:06,339 Weather Foundation, a big weather 1609 00:54:06,340 --> 00:54:08,619 forecasting agency, huge 1610 00:54:08,620 --> 00:54:10,279 computational weather forecasting data, 1611 00:54:10,280 --> 00:54:11,649 the full deal. 1612 00:54:11,650 --> 00:54:12,850 And what he was saying was that 1613 00:54:14,440 --> 00:54:17,439 alongside this, uh, 1614 00:54:17,440 --> 00:54:19,719 uh, growth 1615 00:54:19,720 --> 00:54:21,999 of information available to us is also 1616 00:54:22,000 --> 00:54:23,859 huge and cataclysmic change. 1617 00:54:23,860 --> 00:54:25,929 Climate change is now progressing to 1618 00:54:25,930 --> 00:54:26,919 such an extent. 1619 00:54:26,920 --> 00:54:28,479 The atmosphere is warming to such an 1620 00:54:28,480 --> 00:54:29,860 extent that 1621 00:54:31,120 --> 00:54:33,369 the models that we've built based 1622 00:54:33,370 --> 00:54:35,559 on the 100 plus years of weather 1623 00:54:35,560 --> 00:54:37,719 data we've been gathering, no longer 1624 00:54:37,720 --> 00:54:38,979 apply. 1625 00:54:38,980 --> 00:54:40,299 The weather models are actually getting 1626 00:54:40,300 --> 00:54:43,029 worse because the underlying 1627 00:54:43,030 --> 00:54:45,219 models that generate them no 1628 00:54:45,220 --> 00:54:47,579 longer work so well in 1629 00:54:47,580 --> 00:54:49,389 a rapidly changing climate. 1630 00:54:49,390 --> 00:54:51,489 And we're actually moving towards a place 1631 00:54:51,490 --> 00:54:53,799 where we're going to know less about 1632 00:54:53,800 --> 00:54:56,019 the world than we did previously. 1633 00:54:56,020 --> 00:54:57,549 We're going to be able to predict less, 1634 00:54:57,550 --> 00:55:00,039 and that's going to affect our ability to 1635 00:55:00,040 --> 00:55:01,040 prepare for it. 1636 00:55:05,070 --> 00:55:06,070 This. 1637 00:55:07,910 --> 00:55:09,110 No, not that yes, 1638 00:55:10,430 --> 00:55:11,209 I'm running out of time. 1639 00:55:11,210 --> 00:55:12,329 I want to take a couple of questions, one 1640 00:55:12,330 --> 00:55:14,419 who's got them? And here's what I 1641 00:55:14,420 --> 00:55:15,199 think. 1642 00:55:15,200 --> 00:55:18,379 I do not think the Internet 1643 00:55:18,380 --> 00:55:20,329 and so much of the stuff that we've built 1644 00:55:20,330 --> 00:55:22,759 on top of it is a tool for 1645 00:55:22,760 --> 00:55:24,889 understanding the world as some kind of 1646 00:55:24,890 --> 00:55:27,049 abstract entity that we can 1647 00:55:27,050 --> 00:55:29,239 regard as a static and 1648 00:55:29,240 --> 00:55:31,399 stable. Rather, it is a tool for 1649 00:55:31,400 --> 00:55:33,259 understanding our own ability to 1650 00:55:33,260 --> 00:55:34,969 understand and interact with the world. 1651 00:55:34,970 --> 00:55:37,910 And it's the most extraordinary complex 1652 00:55:39,230 --> 00:55:41,299 and advanced such tool we've built as 1653 00:55:41,300 --> 00:55:42,829 a species. For me, it's a it's a it's a 1654 00:55:42,830 --> 00:55:44,029 cultural tool. It's something that 1655 00:55:44,030 --> 00:55:46,309 emerges like language and writing 1656 00:55:46,310 --> 00:55:47,749 emerged. 1657 00:55:47,750 --> 00:55:49,879 And and what's extraordinary about 1658 00:55:49,880 --> 00:55:52,909 it is that no person or no 1659 00:55:52,910 --> 00:55:55,099 single set of intentions 1660 00:55:55,100 --> 00:55:57,199 created this thing or imagined what we 1661 00:55:57,200 --> 00:55:59,389 might do with it for so its uses. 1662 00:55:59,390 --> 00:56:01,489 It's an unconsciously generated tool for 1663 00:56:01,490 --> 00:56:02,899 unconscious generation. 1664 00:56:02,900 --> 00:56:04,639 And it's this model of complexity. 1665 00:56:04,640 --> 00:56:06,589 It's kind of cloudy cloud. 1666 00:56:06,590 --> 00:56:08,270 And if we are to take our kind of 1667 00:56:09,590 --> 00:56:11,509 technological understanding of the world 1668 00:56:11,510 --> 00:56:13,609 seriously and I feel that we 1669 00:56:13,610 --> 00:56:15,349 must recognize that it's teaching us to 1670 00:56:15,350 --> 00:56:18,139 expect and understand differences 1671 00:56:18,140 --> 00:56:20,149 in the world, to understand cloudiness 1672 00:56:20,150 --> 00:56:22,579 and complexity and not conformity 1673 00:56:22,580 --> 00:56:23,580 and similarity. 1674 00:56:24,590 --> 00:56:25,590 Thank you very much. 1675 00:56:40,270 --> 00:56:42,279 Thank you so much, James. 1676 00:56:42,280 --> 00:56:44,229 We now have a couple more minutes to take 1677 00:56:44,230 --> 00:56:45,230 questions. 1678 00:56:47,150 --> 00:56:49,069 Please use the microphone to show up at 1679 00:56:49,070 --> 00:56:50,449 the microphone so I can see you 1680 00:56:52,890 --> 00:56:54,239 over there afterwards. 1681 00:56:56,360 --> 00:56:57,360 Anyone? 1682 00:57:03,610 --> 00:57:06,009 And right, it's not all Bellator 1683 00:57:06,010 --> 00:57:07,839 is also good. 1684 00:57:07,840 --> 00:57:08,840 OK. 1685 00:57:14,430 --> 00:57:15,359 Are you sure you don't have any 1686 00:57:15,360 --> 00:57:16,360 questions? 1687 00:57:17,580 --> 00:57:18,689 This is your chance. 1688 00:57:22,330 --> 00:57:23,979 I think we're good at that. 1689 00:57:23,980 --> 00:57:25,419 Thank you very much. 1690 00:57:25,420 --> 00:57:26,469 Thank you very much. 1691 00:57:26,470 --> 00:57:27,470 Since Friday.