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/111 Thanks! 1 00:00:09,310 --> 00:00:12,429 To talk and start and yes, 2 00:00:12,430 --> 00:00:14,529 it's about metadata or 3 00:00:14,530 --> 00:00:16,749 against metadata, I think everybody 4 00:00:16,750 --> 00:00:18,359 knows about the metadata by. 5 00:00:18,360 --> 00:00:20,349 So I think I don't have to tell you very 6 00:00:20,350 --> 00:00:21,459 much about the talk. 7 00:00:21,460 --> 00:00:23,620 And so speaking is 8 00:00:24,820 --> 00:00:26,319 Robert Oxhorn. 9 00:00:26,320 --> 00:00:27,609 Yes. 10 00:00:27,610 --> 00:00:28,629 Thank you. 11 00:00:28,630 --> 00:00:29,630 Thanks. 12 00:00:48,610 --> 00:00:50,169 This is what a digital video looks like 13 00:00:50,170 --> 00:00:52,269 without metadata, you are 14 00:00:52,270 --> 00:00:54,309 seeing and hearing a stream of digital 15 00:00:54,310 --> 00:00:56,769 bytes, unlike 16 00:00:56,770 --> 00:00:58,119 a frame of film, which is two 17 00:00:58,120 --> 00:01:00,189 dimensional, no shape in 18 00:01:00,190 --> 00:01:02,649 here is within computer memory. 19 00:01:02,650 --> 00:01:04,869 When we save data to disk, we 20 00:01:04,870 --> 00:01:06,939 must store metadata that allows for 21 00:01:06,940 --> 00:01:08,379 its familiar representation. 22 00:01:09,790 --> 00:01:11,919 In this case, the width and height of 23 00:01:11,920 --> 00:01:12,990 the video are unknown. 24 00:01:15,520 --> 00:01:17,079 And what you are looking at is a user 25 00:01:17,080 --> 00:01:18,579 interface I have developed 26 00:01:21,070 --> 00:01:22,599 for rediscovering the width in the height 27 00:01:22,600 --> 00:01:23,600 of a video. 28 00:01:25,560 --> 00:01:27,279 If I move the mouse to the upper left 29 00:01:27,280 --> 00:01:28,420 hand corner of the screen, 30 00:01:29,440 --> 00:01:31,299 the video plays as if it's two by two. 31 00:01:34,590 --> 00:01:36,809 We can set the width and 32 00:01:36,810 --> 00:01:37,810 the height. 33 00:01:40,450 --> 00:01:41,790 We'll come back to this in a moment. 34 00:02:01,320 --> 00:02:03,599 If we consider that 35 00:02:03,600 --> 00:02:05,459 to be that, which I call signal an 36 00:02:05,460 --> 00:02:08,008 example of data without metadata, 37 00:02:08,009 --> 00:02:11,038 then what I want to show you next noise 38 00:02:11,039 --> 00:02:12,300 is even a lack of data. 39 00:02:14,240 --> 00:02:15,809 The objective here is to cancel out 40 00:02:15,810 --> 00:02:16,810 noise. 41 00:02:22,400 --> 00:02:24,080 I can move the mouse to position 42 00:02:26,060 --> 00:02:28,189 a layer of antennas which is 43 00:02:28,190 --> 00:02:29,389 identical to noise, 44 00:02:31,490 --> 00:02:33,619 but I have no hints as to where they will 45 00:02:33,620 --> 00:02:34,620 meet 46 00:02:36,140 --> 00:02:37,140 if I zoom in. 47 00:02:39,820 --> 00:02:40,820 A little more hope 48 00:02:42,100 --> 00:02:43,300 and see. 49 00:02:53,020 --> 00:02:54,340 Is there sound, by the way, 50 00:02:55,600 --> 00:02:57,399 or is that something that stopped? 51 00:03:02,490 --> 00:03:04,349 Here I start to see the patterns emerging 52 00:03:04,350 --> 00:03:06,419 as the anti noise meets noise. 53 00:03:09,810 --> 00:03:10,810 And then there's nothing 54 00:03:13,620 --> 00:03:14,819 there's nothing that can be found in 55 00:03:14,820 --> 00:03:16,650 noise, the only pattern is itself. 56 00:03:18,330 --> 00:03:19,559 You might have noticed some visual 57 00:03:19,560 --> 00:03:21,779 similarity at first between noise and 58 00:03:21,780 --> 00:03:22,780 signal. 59 00:03:25,510 --> 00:03:27,669 Without metadata, it can be hard to tell 60 00:03:27,670 --> 00:03:28,670 the two apart. 61 00:03:34,390 --> 00:03:35,499 I'm going to come back to this for a 62 00:03:35,500 --> 00:03:36,759 second and try to decode it 63 00:03:37,820 --> 00:03:39,120 sometimes a little difficult. 64 00:03:45,570 --> 00:03:47,169 And moving the mouse around to safety 65 00:03:47,170 --> 00:03:48,170 with an. 66 00:04:37,430 --> 00:04:39,919 The video that I'm trying to decode 67 00:04:39,920 --> 00:04:42,289 was released by WikiLeaks 68 00:04:42,290 --> 00:04:43,850 under the name Collateral Murder. 69 00:04:49,980 --> 00:04:52,349 I developed this interface between 70 00:04:52,350 --> 00:04:54,419 2010 and 2011 in conversation with 71 00:04:54,420 --> 00:04:56,729 my friend's death, inspired by a remark 72 00:04:56,730 --> 00:04:58,919 Julian Assange made that 73 00:04:58,920 --> 00:05:00,659 his organization had decrypted the video 74 00:05:00,660 --> 00:05:01,860 prior to its release. 75 00:05:03,480 --> 00:05:05,999 We wanted to give others the opportunity 76 00:05:06,000 --> 00:05:08,249 to decrypt the video, to actively 77 00:05:08,250 --> 00:05:10,709 take part in overcoming 78 00:05:10,710 --> 00:05:12,989 the formidable hurdles we face to bear 79 00:05:12,990 --> 00:05:13,990 witness. 80 00:05:15,140 --> 00:05:16,979 From what I can tell, the technical 81 00:05:16,980 --> 00:05:18,419 reality of Assange is playing this 82 00:05:18,420 --> 00:05:19,420 controversial. 83 00:05:21,700 --> 00:05:23,909 From the test to the 84 00:05:23,910 --> 00:05:26,339 trial of Chelsea Manning, 85 00:05:26,340 --> 00:05:28,379 there's some indication that the soldier 86 00:05:28,380 --> 00:05:30,449 had access to an unencrypted 87 00:05:30,450 --> 00:05:31,569 version of this video. 88 00:05:35,630 --> 00:05:36,889 Still, I think there's something to what 89 00:05:36,890 --> 00:05:38,989 Assange says, said, because 90 00:05:38,990 --> 00:05:41,269 even if the organization didn't perform 91 00:05:41,270 --> 00:05:43,339 magic crypto wizardry to 92 00:05:43,340 --> 00:05:45,559 bring the video to our attention, if 93 00:05:45,560 --> 00:05:47,029 he wasn't using this interface or 94 00:05:47,030 --> 00:05:48,260 anything remotely like it. 95 00:05:52,260 --> 00:05:54,749 He did decrypt it in a way, 96 00:05:54,750 --> 00:05:56,699 because the release of the video, 97 00:05:58,620 --> 00:06:00,119 starting with its title as Collateral 98 00:06:00,120 --> 00:06:02,249 Murder, was not merely a release of 99 00:06:02,250 --> 00:06:05,309 raw material, but also metadata. 100 00:06:05,310 --> 00:06:07,079 The video was not decrypted to find it. 101 00:06:07,080 --> 00:06:09,179 Machine readable metadata, the 102 00:06:09,180 --> 00:06:11,549 crypto protocol, privity, audio video 103 00:06:11,550 --> 00:06:13,439 codec with height and frame rate. 104 00:06:13,440 --> 00:06:14,879 But it was decrypted to make human 105 00:06:14,880 --> 00:06:16,410 readable claims about the video. 106 00:06:22,130 --> 00:06:24,819 I think I'm about to give up on trying to 107 00:06:24,820 --> 00:06:25,820 solve it. 108 00:06:41,830 --> 00:06:43,389 The title of this talk is Against 109 00:06:43,390 --> 00:06:45,759 Metadata, submitted in the early midst 110 00:06:45,760 --> 00:06:48,309 of dragnet surveillance disclosures, 111 00:06:48,310 --> 00:06:49,779 but perhaps after a more sober 112 00:06:49,780 --> 00:06:51,879 consideration, it should have been 113 00:06:51,880 --> 00:06:54,159 called beyond metadata or towards the 114 00:06:54,160 --> 00:06:56,079 new metadata or some such. 115 00:06:58,750 --> 00:07:01,179 The summary of Snowden as apparently 116 00:07:01,180 --> 00:07:03,279 here is known evoked 117 00:07:03,280 --> 00:07:05,349 in me a sort of double sadness that all 118 00:07:05,350 --> 00:07:07,449 this metadata had leaked and 119 00:07:07,450 --> 00:07:09,309 perversely, that the dragnet ran so 120 00:07:09,310 --> 00:07:10,310 shallow. 121 00:07:11,620 --> 00:07:13,659 Metadata is too banal and frankly 122 00:07:13,660 --> 00:07:16,059 necessary to antagonize. 123 00:07:16,060 --> 00:07:18,219 But a plea stop designing 124 00:07:18,220 --> 00:07:20,889 software that assumes the prior existence 125 00:07:20,890 --> 00:07:21,890 of metadata. 126 00:07:23,260 --> 00:07:25,029 I would like to split metadata into three 127 00:07:25,030 --> 00:07:26,649 categories. 128 00:07:26,650 --> 00:07:28,929 First, there is machine metadata 129 00:07:28,930 --> 00:07:30,729 designed for the computer to interpret 130 00:07:30,730 --> 00:07:32,619 out of our sight, such as the width and 131 00:07:32,620 --> 00:07:35,049 height of a video stream, its frame rate 132 00:07:35,050 --> 00:07:37,149 codecs, etc.. 133 00:07:37,150 --> 00:07:39,099 Then we have metadata bases full of 134 00:07:39,100 --> 00:07:41,169 relational columns that, more so 135 00:07:41,170 --> 00:07:43,179 than the original, are suited to rigid 136 00:07:43,180 --> 00:07:45,459 database systems and which are then used 137 00:07:45,460 --> 00:07:48,399 to stand in place of the original data, 138 00:07:48,400 --> 00:07:50,829 such as a book, cover titles 139 00:07:50,830 --> 00:07:53,859 and names of all sorts icons, timestamps, 140 00:07:53,860 --> 00:07:56,199 GPS coordinates and so on. 141 00:07:56,200 --> 00:07:58,359 Finally, the last sort of metadata I wish 142 00:07:58,360 --> 00:08:00,459 to consider is interpretive metadata, 143 00:08:00,460 --> 00:08:02,169 which is a layer of data describing what 144 00:08:02,170 --> 00:08:03,170 we make of the data. 145 00:08:20,060 --> 00:08:21,949 Compression is the basis and enabling 146 00:08:21,950 --> 00:08:24,049 force of networked media, and as 147 00:08:24,050 --> 00:08:26,029 with all machine metadata, it is deemed 148 00:08:26,030 --> 00:08:27,709 successful when it is invisible. 149 00:08:28,790 --> 00:08:30,379 This is an image which is represented 150 00:08:30,380 --> 00:08:32,359 digitally as an array of pixel intensity 151 00:08:32,360 --> 00:08:33,469 values. 152 00:08:33,470 --> 00:08:35,450 I can mix wires and turn up the volume 153 00:08:36,770 --> 00:08:38,959 so you can hear these values 154 00:08:38,960 --> 00:08:39,960 played back. 155 00:08:41,030 --> 00:08:42,769 Compression inherently involves a 156 00:08:42,770 --> 00:08:44,749 determination of what information is most 157 00:08:44,750 --> 00:08:47,929 important and a purge of the rest. 158 00:08:47,930 --> 00:08:49,819 For example, compress images tend to 159 00:08:49,820 --> 00:08:51,949 assume a basic continuity that 160 00:08:51,950 --> 00:08:54,169 pixels are similar to their neighbors 161 00:08:54,170 --> 00:08:55,909 and therefore operate on spatial 162 00:08:55,910 --> 00:08:57,019 frequencies of pixels. 163 00:08:58,310 --> 00:09:00,109 So while we perceive this image as a two 164 00:09:00,110 --> 00:09:02,689 degree grid, the computer processes 165 00:09:02,690 --> 00:09:04,460 it a little bit more like this. 166 00:09:07,490 --> 00:09:09,199 And I can take these frequencies and one 167 00:09:09,200 --> 00:09:11,779 at a time start removing them 168 00:09:11,780 --> 00:09:13,999 like you can hear, I'm 169 00:09:14,000 --> 00:09:15,889 removing the pixel, the frequencies that 170 00:09:15,890 --> 00:09:17,959 are the least salient 171 00:09:17,960 --> 00:09:19,909 and I can flip back to the image and we 172 00:09:19,910 --> 00:09:22,219 see that it's more or less unchanged. 173 00:09:22,220 --> 00:09:23,929 This is with 50 percent of that frequency 174 00:09:23,930 --> 00:09:25,369 is removed, which you may be able to 175 00:09:25,370 --> 00:09:26,359 hear. 176 00:09:26,360 --> 00:09:27,360 I can keep going. 177 00:09:31,440 --> 00:09:33,019 About seventy five percent, 178 00:09:35,230 --> 00:09:37,499 nine days, and 179 00:09:37,500 --> 00:09:39,599 finally all of the frequencies 180 00:09:39,600 --> 00:09:41,819 are removed, I can add them back 181 00:09:41,820 --> 00:09:42,820 one at a time. 182 00:10:01,880 --> 00:10:03,739 And we can see the images start to become 183 00:10:03,740 --> 00:10:04,840 recognizable again, 184 00:10:06,680 --> 00:10:08,179 what would image compression looked like 185 00:10:08,180 --> 00:10:10,669 if it was oriented towards humans 186 00:10:10,670 --> 00:10:11,670 and not machines? 187 00:10:24,700 --> 00:10:26,169 These are a series of portraits, 188 00:10:27,220 --> 00:10:28,220 and I've 189 00:10:32,020 --> 00:10:34,179 given myself a parameter over each image, 190 00:10:34,180 --> 00:10:35,259 so I may compress it 191 00:10:37,240 --> 00:10:38,799 by horribly abusing the math 192 00:10:39,880 --> 00:10:40,990 from some carving 193 00:10:42,130 --> 00:10:44,679 popularized by a paper on contemporary 194 00:10:44,680 --> 00:10:46,839 imagery resizing, it was designed 195 00:10:46,840 --> 00:10:49,119 to invisibly change the aspect 196 00:10:49,120 --> 00:10:51,249 ratio of images for different 197 00:10:51,250 --> 00:10:52,570 sizes of displays. 198 00:10:55,780 --> 00:10:57,549 I hope that by dialing up all of these 199 00:10:57,550 --> 00:10:58,550 images at once. 200 00:11:02,500 --> 00:11:04,239 I can get a few laughs out of the 201 00:11:04,240 --> 00:11:05,240 audience. 202 00:11:07,530 --> 00:11:09,299 I see the ideal of metadata as 203 00:11:09,300 --> 00:11:11,399 encompassing the meaning of that data 204 00:11:11,400 --> 00:11:13,469 and compression as a proposal for that 205 00:11:13,470 --> 00:11:15,689 meaning, if you laughed at these 206 00:11:15,690 --> 00:11:17,669 images, perhaps it is because you got 207 00:11:17,670 --> 00:11:19,829 more than you expected for them, that as 208 00:11:19,830 --> 00:11:21,599 a caricature, the compression functioned 209 00:11:21,600 --> 00:11:23,399 as an exaggeration of distinctive 210 00:11:23,400 --> 00:11:24,400 characteristics. 211 00:11:25,710 --> 00:11:27,209 In nineteen fifty, the pioneer of 212 00:11:27,210 --> 00:11:29,549 information theory, Claude Shannon, 213 00:11:29,550 --> 00:11:31,559 wrote, The meaning of a message is 214 00:11:31,560 --> 00:11:33,569 generally irrelevant. 215 00:11:33,570 --> 00:11:35,159 When I read that at first I took issue 216 00:11:35,160 --> 00:11:37,529 with it. Shannon defines entropy 217 00:11:37,530 --> 00:11:39,899 or information relative to the average 218 00:11:39,900 --> 00:11:41,939 or expected message. 219 00:11:41,940 --> 00:11:44,219 Information is surprise, Shannon 220 00:11:44,220 --> 00:11:46,439 writes, and the encoding of a normal 221 00:11:46,440 --> 00:11:48,719 baseline into the superstructure of file 222 00:11:48,720 --> 00:11:50,999 formats and media codecs seemed 223 00:11:51,000 --> 00:11:53,279 a reactionary bias towards mediocrity. 224 00:11:54,300 --> 00:11:56,849 From telegrammed to t nine codes. 225 00:11:56,850 --> 00:11:58,859 A machine may not be able to comprehend 226 00:11:58,860 --> 00:12:00,959 meaning, but it can be programed to let 227 00:12:00,960 --> 00:12:02,849 some pass through a slit. 228 00:12:02,850 --> 00:12:04,409 And this is another side to Shannon's 229 00:12:04,410 --> 00:12:06,149 statement that perhaps I missed the first 230 00:12:06,150 --> 00:12:07,229 time around. 231 00:12:07,230 --> 00:12:09,569 Meaning need not be encapsulated within 232 00:12:09,570 --> 00:12:11,669 a packet for it to be carried through the 233 00:12:11,670 --> 00:12:12,670 medium. 234 00:12:33,570 --> 00:12:36,569 For temporal media, sound and video 235 00:12:36,570 --> 00:12:38,639 codecs have given a greater and greater 236 00:12:38,640 --> 00:12:40,709 instantaneous fidelity, but leave 237 00:12:40,710 --> 00:12:43,289 us with few techniques to skim, seek 238 00:12:43,290 --> 00:12:45,479 and survey, the timeline 239 00:12:45,480 --> 00:12:47,519 is a fixture of digital non-linear 240 00:12:47,520 --> 00:12:49,709 editors, but rarely in our players is 241 00:12:49,710 --> 00:12:51,299 it much more than a line. 242 00:12:51,300 --> 00:12:53,489 This is 60 seconds of video represented 243 00:12:53,490 --> 00:12:55,769 with 12 thumbnails, each spaced five 244 00:12:55,770 --> 00:12:56,770 seconds apart. 245 00:13:00,360 --> 00:13:02,369 This is the same 60 seconds, but now with 246 00:13:02,370 --> 00:13:04,109 twenty four thumbnails, two and a half 247 00:13:04,110 --> 00:13:07,289 seconds apart, left, half, 248 00:13:07,290 --> 00:13:08,290 right half. 249 00:13:09,720 --> 00:13:11,759 And when I keep cropping the images more 250 00:13:11,760 --> 00:13:14,039 and more down to a single 251 00:13:14,040 --> 00:13:16,199 pixel, all of a sudden 252 00:13:16,200 --> 00:13:18,299 form reemerges within 253 00:13:18,300 --> 00:13:20,369 the image. People in cars are visible 254 00:13:20,370 --> 00:13:21,370 again. 255 00:13:22,290 --> 00:13:24,149 Movement is frozen through the column of 256 00:13:24,150 --> 00:13:26,399 pixels I choose I take 257 00:13:26,400 --> 00:13:27,960 in no matter which column that is. 258 00:13:29,310 --> 00:13:31,469 I can look at the next 60 seconds 259 00:13:31,470 --> 00:13:33,599 and see a similar effect 260 00:13:33,600 --> 00:13:34,979 or the next 60 seconds. 261 00:13:36,510 --> 00:13:38,759 Now we see a parabola and 262 00:13:38,760 --> 00:13:40,919 when I uncropped the images, we see a 263 00:13:40,920 --> 00:13:41,920 series of them. 264 00:13:42,750 --> 00:13:44,490 I'm interested in a point about here 265 00:13:45,540 --> 00:13:48,239 where you see the form, you see the icon, 266 00:13:48,240 --> 00:13:50,339 the clock, the buildings, but you also 267 00:13:50,340 --> 00:13:52,079 see how they relate to time and how they 268 00:13:52,080 --> 00:13:53,969 change the part. 269 00:13:53,970 --> 00:13:54,970 And the whole 270 00:13:57,600 --> 00:13:59,699 the timeline can function as 271 00:13:59,700 --> 00:14:00,809 a graphic metadata 272 00:14:03,360 --> 00:14:05,009 to communicate the contents of video 273 00:14:05,010 --> 00:14:07,079 without, in fact, having any idea of 274 00:14:07,080 --> 00:14:08,080 them. 275 00:14:14,530 --> 00:14:16,689 Here's the timelines index, 15 276 00:14:16,690 --> 00:14:17,690 films by Cut 277 00:14:18,760 --> 00:14:19,959 at different phase. 278 00:14:25,820 --> 00:14:27,559 And I can play the video through these 279 00:14:27,560 --> 00:14:28,560 timelines. 280 00:14:34,630 --> 00:14:35,980 Using them as index. 281 00:14:58,420 --> 00:15:00,669 These timelines eventually turned 282 00:15:00,670 --> 00:15:03,129 into a Web based documentary 283 00:15:03,130 --> 00:15:05,599 I did in collaboration with the filmmaker 284 00:15:05,600 --> 00:15:06,600 L7. 285 00:15:09,460 --> 00:15:11,619 Here we have metadata as the 286 00:15:11,620 --> 00:15:12,620 film. 287 00:15:18,440 --> 00:15:20,059 Tags serve as entry points to the 288 00:15:20,060 --> 00:15:21,060 material. 289 00:15:28,620 --> 00:15:30,719 Nobody can say that 290 00:15:30,720 --> 00:15:33,569 it is to blame also, do you 291 00:15:33,570 --> 00:15:35,090 see that one 292 00:15:37,110 --> 00:15:38,110 born 293 00:15:40,230 --> 00:15:42,490 again we have these scandals 294 00:15:43,950 --> 00:15:45,230 usually created more was 295 00:15:46,540 --> 00:15:48,899 on the tags, allow us to pivot 296 00:15:48,900 --> 00:15:50,819 to change how we're seeing something 297 00:15:50,820 --> 00:15:53,099 without it stopping so bad 298 00:15:53,100 --> 00:15:54,100 that I haven't, you 299 00:15:55,390 --> 00:15:56,390 know. 300 00:15:58,840 --> 00:16:00,949 In the middle of the 301 00:16:00,950 --> 00:16:03,009 day, and I'm older than me, 302 00:16:03,010 --> 00:16:04,760 but I am little. 303 00:16:13,150 --> 00:16:14,919 Metadata as authorship is only the 304 00:16:14,920 --> 00:16:15,920 beginning, 305 00:16:17,260 --> 00:16:19,839 for if this is the real time generation, 306 00:16:19,840 --> 00:16:22,089 real must be intended as in reality 307 00:16:22,090 --> 00:16:23,090 TV. 308 00:16:32,540 --> 00:16:34,579 The rise of portable MP three players 309 00:16:34,580 --> 00:16:36,379 gave us the opportunity for our lives to 310 00:16:36,380 --> 00:16:38,599 have soundtrack's and the smartphone, 311 00:16:38,600 --> 00:16:40,879 NSA spyglasses, etc, all 312 00:16:40,880 --> 00:16:42,859 give us thickening metadata layer to 313 00:16:42,860 --> 00:16:44,699 lived experience. 314 00:16:44,700 --> 00:16:46,579 We are making history as we are living 315 00:16:46,580 --> 00:16:48,289 it, which I mean less than a 316 00:16:48,290 --> 00:16:49,879 congratulatory reference to our 317 00:16:49,880 --> 00:16:51,229 significance. 318 00:16:51,230 --> 00:16:53,689 But simply pointing to the accumulation 319 00:16:53,690 --> 00:16:55,849 of documents and data is that we or 320 00:16:55,850 --> 00:16:57,769 someone something must sort through. 321 00:17:00,140 --> 00:17:02,269 Here I am live coding a video editor 322 00:17:02,270 --> 00:17:04,338 to organize a screen, tests of myself, 323 00:17:04,339 --> 00:17:06,019 life coding, a video editor. 324 00:17:12,400 --> 00:17:14,469 It is the classic Burhenn problem of 325 00:17:14,470 --> 00:17:16,539 the map, and the territory for every 326 00:17:16,540 --> 00:17:19,449 attempt at curation is also recorded 327 00:17:19,450 --> 00:17:20,470 and reinvested. 328 00:17:40,250 --> 00:17:41,719 Bernays wrote about the empire's 329 00:17:41,720 --> 00:17:43,699 fanatical attempt to create a map so 330 00:17:43,700 --> 00:17:45,289 detailed it ended up as large as the 331 00:17:45,290 --> 00:17:47,449 space it sought to describe the 332 00:17:47,450 --> 00:17:49,069 map occluding and suffocating the 333 00:17:49,070 --> 00:17:50,539 territory. 334 00:17:50,540 --> 00:17:52,249 But that was before the advent of slippy 335 00:17:52,250 --> 00:17:54,169 maps, before level of detail 336 00:17:54,170 --> 00:17:55,939 distillations. 337 00:17:55,940 --> 00:17:57,569 We had the present moment. 338 00:17:57,570 --> 00:17:59,239 And while a map of the future is hard to 339 00:17:59,240 --> 00:18:01,309 come by, we can control our view of 340 00:18:01,310 --> 00:18:03,170 the past with a fluid zoom control. 341 00:18:22,100 --> 00:18:23,629 Try to work your head around the passage 342 00:18:23,630 --> 00:18:25,759 of time here in 343 00:18:25,760 --> 00:18:28,009 the video metadata, the traps us neither 344 00:18:28,010 --> 00:18:30,109 in the present or the past. 345 00:18:33,120 --> 00:18:35,039 Applied back to the first time lines, I 346 00:18:35,040 --> 00:18:37,049 showed you those from John Smith's 347 00:18:37,050 --> 00:18:39,059 brilliant nineteen seventy six short 348 00:18:39,060 --> 00:18:41,519 film, The Girl Chewing Gum, itself 349 00:18:41,520 --> 00:18:44,369 a biting critique of cinematic authority. 350 00:18:44,370 --> 00:18:45,869 We have an experimental film 351 00:18:47,040 --> 00:18:48,330 I made called Chewing. 352 00:18:57,700 --> 00:18:59,979 Notice how time passes and also 353 00:18:59,980 --> 00:19:00,980 how it doesn't 354 00:19:02,350 --> 00:19:04,689 how characters are endlessly looped into 355 00:19:04,690 --> 00:19:05,690 themselves. 356 00:19:09,640 --> 00:19:11,889 So we can design video codecs that give 357 00:19:11,890 --> 00:19:14,019 rather than remove context, 358 00:19:14,020 --> 00:19:16,329 all of the massive codecs is valuable. 359 00:19:16,330 --> 00:19:17,589 And I hope that some of you who 360 00:19:17,590 --> 00:19:19,179 intuitively understand it better than 361 00:19:19,180 --> 00:19:21,369 myself will work on opening up 362 00:19:21,370 --> 00:19:23,739 codecs to our perception 363 00:19:23,740 --> 00:19:25,029 to show you what I mean in a different 364 00:19:25,030 --> 00:19:27,039 way. Consider the matter of a video 365 00:19:27,040 --> 00:19:28,040 stabilizer. 366 00:19:30,580 --> 00:19:32,589 Put simply, a stabilizer deforms each 367 00:19:32,590 --> 00:19:34,749 frame to the previous, and when 368 00:19:34,750 --> 00:19:36,519 engineers design stabilizers, 369 00:19:37,780 --> 00:19:40,329 they zoom in to hide the edges that are 370 00:19:40,330 --> 00:19:41,330 irregular. 371 00:19:46,130 --> 00:19:48,019 But what if instead of zooming in, we can 372 00:19:48,020 --> 00:19:49,020 zoom out? 373 00:20:03,530 --> 00:20:06,709 Now we have context, 374 00:20:06,710 --> 00:20:08,749 we have what is no longer seen, I call 375 00:20:08,750 --> 00:20:10,909 this technique periphery in admiration of 376 00:20:10,910 --> 00:20:13,069 our perceptual system, which gives us 377 00:20:13,070 --> 00:20:14,710 so much out of sight context, 378 00:20:15,920 --> 00:20:17,689 I hope I've suggested a few ways the 379 00:20:17,690 --> 00:20:19,609 machine metadata can be designed to 380 00:20:19,610 --> 00:20:21,919 reveal rather than obscure and 381 00:20:21,920 --> 00:20:24,619 have the inevitable metadata as we form 382 00:20:24,620 --> 00:20:26,299 can empower our curation and 383 00:20:26,300 --> 00:20:28,459 comprehension rather than functioning 384 00:20:28,460 --> 00:20:30,079 as advertising and entertainment. 385 00:20:32,450 --> 00:20:34,549 The most practical software that 386 00:20:34,550 --> 00:20:36,829 I've been designing for this 387 00:20:36,830 --> 00:20:38,119 is called interface. 388 00:20:38,120 --> 00:20:40,369 I have very limited time, but I show 389 00:20:40,370 --> 00:20:41,370 you very quickly, 390 00:20:42,800 --> 00:20:44,869 um, a 391 00:20:44,870 --> 00:20:47,029 little screenshot, a little screen of 392 00:20:47,030 --> 00:20:48,919 a project I did in collaboration with a 393 00:20:48,920 --> 00:20:51,169 conference suitably enough for my 394 00:20:51,170 --> 00:20:52,910 metastasize about video 395 00:20:54,290 --> 00:20:55,999 that perhaps instead of just 396 00:20:56,000 --> 00:20:58,279 understanding that as reaction, 397 00:20:58,280 --> 00:21:00,469 there is the entire conference which 398 00:21:00,470 --> 00:21:02,909 is visible through a zoomed assumable 399 00:21:02,910 --> 00:21:03,910 timelines. 400 00:21:05,030 --> 00:21:06,619 And then there's a series of footnotes 401 00:21:06,620 --> 00:21:08,329 that provide an alternate mode of 402 00:21:08,330 --> 00:21:10,339 navigation. He talks about 403 00:21:12,200 --> 00:21:14,539 it back in his 404 00:21:14,540 --> 00:21:16,640 dorm room in Georgia while. 405 00:21:20,360 --> 00:21:21,769 There's a theory, if you want to change 406 00:21:21,770 --> 00:21:24,049 the world, that you have to do it by 407 00:21:24,050 --> 00:21:26,569 helping people share cat videos. 408 00:21:26,570 --> 00:21:28,069 I don't quite know about that, but I'll 409 00:21:28,070 --> 00:21:29,929 show you some thing a little bit less 410 00:21:29,930 --> 00:21:32,299 serious fed into this sort of machine. 411 00:21:44,020 --> 00:21:45,020 See how it does 412 00:21:46,350 --> 00:21:48,529 and how 413 00:21:48,530 --> 00:21:49,989 maybe we should take questions as you 414 00:21:49,990 --> 00:21:51,819 look at this snorkeling video had some 415 00:21:51,820 --> 00:21:53,469 other things, but I have absolutely no 416 00:21:53,470 --> 00:21:54,729 time anyone. 417 00:22:07,410 --> 00:22:08,909 If there are no questions, you have five 418 00:22:08,910 --> 00:22:09,910 minutes. 419 00:22:13,030 --> 00:22:14,809 This is called this 420 00:22:16,140 --> 00:22:18,449 and the software is codas 421 00:22:18,450 --> 00:22:20,279 on README not button OMG. 422 00:22:20,280 --> 00:22:22,019 There have been a quite a few iterations, 423 00:22:22,020 --> 00:22:24,240 but now the latest is interlaced for 424 00:22:26,280 --> 00:22:27,749 Alpha Grid. 425 00:22:27,750 --> 00:22:28,750 Be patient. 426 00:22:30,120 --> 00:22:31,120 That's one question. 427 00:22:32,490 --> 00:22:34,739 I guess I'm taking your challenge 428 00:22:34,740 --> 00:22:36,869 of curation quite literally. 429 00:22:36,870 --> 00:22:39,539 I'm an archivist with a 430 00:22:39,540 --> 00:22:41,669 collection of archival films 431 00:22:41,670 --> 00:22:43,019 and something like this. 432 00:22:43,020 --> 00:22:45,449 On the one hand, which provides 433 00:22:45,450 --> 00:22:47,969 an overview of the entire contents 434 00:22:47,970 --> 00:22:50,519 of a film, would be incredibly 435 00:22:50,520 --> 00:22:53,639 useful as a 436 00:22:53,640 --> 00:22:55,649 tool, as a finding, a way to our 437 00:22:55,650 --> 00:22:58,019 collections. But on the other hand, 438 00:22:58,020 --> 00:23:00,119 if it is video, 439 00:23:00,120 --> 00:23:02,699 there's we have to design. 440 00:23:02,700 --> 00:23:04,859 What we put out on the Web 441 00:23:04,860 --> 00:23:07,139 is a finding aid for the absolute 442 00:23:07,140 --> 00:23:09,509 lowest common denominator of 443 00:23:09,510 --> 00:23:12,299 technology of users worldwide. 444 00:23:13,860 --> 00:23:15,749 Is there a tension there? 445 00:23:15,750 --> 00:23:17,999 And how would somebody like 446 00:23:18,000 --> 00:23:19,799 how would record keepers like me 447 00:23:21,270 --> 00:23:23,759 sort of smooth over that tension 448 00:23:23,760 --> 00:23:25,679 in terms of lowest common denominator? 449 00:23:25,680 --> 00:23:27,029 You're looking at a Web browser 450 00:23:28,140 --> 00:23:29,759 and pushing the limits of the browser a 451 00:23:29,760 --> 00:23:31,859 little bit, but it's HTML5. 452 00:23:31,860 --> 00:23:33,029 It's all open source. 453 00:23:33,030 --> 00:23:35,369 There's no lock in and there's 454 00:23:35,370 --> 00:23:37,469 no barriers to free implementations 455 00:23:37,470 --> 00:23:38,879 that would provide access. 456 00:23:38,880 --> 00:23:40,619 Usability is still something I'm working 457 00:23:40,620 --> 00:23:43,319 on. On the other end, yeah, 458 00:23:43,320 --> 00:23:45,209 I guess I'm thinking more of bandwidth 459 00:23:45,210 --> 00:23:48,599 because something that's sort of a static 460 00:23:48,600 --> 00:23:51,479 series of images 461 00:23:51,480 --> 00:23:54,059 takes up a lot less space, 462 00:23:54,060 --> 00:23:56,639 really, than than sort of here is 463 00:23:56,640 --> 00:23:58,529 on the other hand, having a visual index 464 00:23:58,530 --> 00:24:00,329 lets you watch the part of the archive. 465 00:24:00,330 --> 00:24:01,769 You're interested in downloading the 466 00:24:01,770 --> 00:24:03,959 entire file exactly as 467 00:24:03,960 --> 00:24:05,759 more bandwidth. 468 00:24:05,760 --> 00:24:07,859 I did a lot of optimization to try 469 00:24:07,860 --> 00:24:09,989 to make this like vaguely palatable 470 00:24:09,990 --> 00:24:11,339 over a network connection. 471 00:24:13,950 --> 00:24:16,829 But I mean, these are tiny 472 00:24:16,830 --> 00:24:17,789 little videos. 473 00:24:17,790 --> 00:24:18,790 They're 474 00:24:20,670 --> 00:24:21,929 just indexes. 475 00:24:21,930 --> 00:24:23,039 It's like a search index. 476 00:24:23,040 --> 00:24:24,040 But for video. 477 00:24:26,780 --> 00:24:29,209 And I don't know if I've answered any of 478 00:24:29,210 --> 00:24:31,309 you, please maybe ask 479 00:24:31,310 --> 00:24:32,569 the most important again. 480 00:24:32,570 --> 00:24:34,669 Yeah, actually I can't 481 00:24:34,670 --> 00:24:35,629 remember what it is. 482 00:24:35,630 --> 00:24:37,100 I'm going to have to 483 00:24:38,210 --> 00:24:39,739 basically think about it. 484 00:24:39,740 --> 00:24:41,689 But then we have another question out 485 00:24:41,690 --> 00:24:43,849 there. If you have a question, can you go 486 00:24:43,850 --> 00:24:45,949 to the microphone and 487 00:24:45,950 --> 00:24:46,950 my phone number one. 488 00:24:52,030 --> 00:24:54,039 I think that was really great. 489 00:24:54,040 --> 00:24:55,599 Unfortunately, I don't have the pleasure 490 00:24:55,600 --> 00:24:57,729 of working with video as my main 491 00:24:57,730 --> 00:25:00,129 source of data today, and it's 492 00:25:00,130 --> 00:25:02,499 much more dry a context of numbers and 493 00:25:02,500 --> 00:25:04,029 context. 494 00:25:04,030 --> 00:25:06,279 And I can see some 495 00:25:06,280 --> 00:25:09,219 similarities in what you've done with, 496 00:25:09,220 --> 00:25:11,559 you know, spear cons and spin 497 00:25:11,560 --> 00:25:12,759 Dex's or however you supposed to 498 00:25:12,760 --> 00:25:14,859 pronounce it with sonification and 499 00:25:14,860 --> 00:25:15,789 some visualization. 500 00:25:15,790 --> 00:25:16,790 But I want to have you. 501 00:25:18,250 --> 00:25:19,720 Yes. Well, I'm just going to ask you to 502 00:25:20,980 --> 00:25:21,980 do that. Yeah, 503 00:25:26,050 --> 00:25:27,369 I had where I wanted to show, 504 00:25:28,870 --> 00:25:30,069 but I've tried to apply the same 505 00:25:30,070 --> 00:25:32,379 paradigms to dump's of unsorted 506 00:25:32,380 --> 00:25:34,509 freedom of Information requests and to 507 00:25:34,510 --> 00:25:37,299 sound with spectrograms and provide 508 00:25:37,300 --> 00:25:39,549 ways of not linearly 509 00:25:39,550 --> 00:25:41,799 exploring all sorts of different kinds 510 00:25:41,800 --> 00:25:43,599 of data that doesn't have a clear 511 00:25:43,600 --> 00:25:45,999 archival canonical metadata available 512 00:25:46,000 --> 00:25:47,000 to it. 513 00:25:51,050 --> 00:25:53,030 If there's no other question, I think 514 00:25:54,530 --> 00:25:56,509 it's to say thank you again.