1 00:00:00,000 --> 00:00:02,750 (dramatic music) 2 00:00:03,900 --> 00:00:08,690 What if we asked scientists 3 00:00:11,430 --> 00:00:16,123 to come up with something truly unique? 4 00:00:16,123 --> 00:00:21,123 The first piece of musical theater generated by computers. 5 00:00:21,709 --> 00:00:25,694 ♪ The hills are alive with the sound of ♪ 6 00:00:25,694 --> 00:00:27,927 (cacophonous electronic music) 7 00:00:27,927 --> 00:00:31,463 We've called this endeavor the Experiment. 8 00:00:33,170 --> 00:00:34,650 I thought it was bonkers. 9 00:00:34,650 --> 00:00:36,475 For us, it was wonderful. 10 00:00:36,475 --> 00:00:38,330 Our international team of scientists 11 00:00:38,330 --> 00:00:40,130 conducts a groundbreaking study 12 00:00:40,130 --> 00:00:42,430 into what makes a hit musical. 13 00:00:42,430 --> 00:00:43,594 How much death? 14 00:00:43,594 --> 00:00:44,490 (laughing) 15 00:00:44,490 --> 00:00:46,030 And then makes one. 16 00:00:46,030 --> 00:00:48,680 I've just generated a show tune by my program. 17 00:00:48,680 --> 00:00:51,270 I call my program Android Lloyd Webber. 18 00:00:51,270 --> 00:00:52,740 Still gibberish. 19 00:00:52,740 --> 00:00:54,530 This machine is talking to a machine 20 00:00:54,530 --> 00:00:58,321 in Madrid, in Cambridge, and in Slovenia now. 21 00:00:58,321 --> 00:00:59,370 (group vocalizing) 22 00:00:59,370 --> 00:01:01,310 A top cast and creative team 23 00:01:01,310 --> 00:01:03,650 will stage it in London's West End. 24 00:01:03,650 --> 00:01:05,136 Standing by then, please. 25 00:01:05,136 --> 00:01:06,820 ♪ Soggy and ♪ 26 00:01:06,820 --> 00:01:07,900 This is a meeting 27 00:01:07,900 --> 00:01:10,127 of science and art as never before. 28 00:01:10,127 --> 00:01:13,564 And if the show's ghastly, it's not really my fault. 29 00:01:13,564 --> 00:01:16,014 Even if it fails, it's necessary to experiment. 30 00:01:17,010 --> 00:01:18,480 And bridging the two 31 00:01:18,480 --> 00:01:21,180 are Benjamin Till and Nathan Taylor, 32 00:01:21,180 --> 00:01:23,907 musical theater writers who, on this occasion, 33 00:01:23,907 --> 00:01:27,163 will take dictation from the computers. 34 00:01:28,240 --> 00:01:30,860 Why would you want a computer to write a musical? 35 00:01:30,860 --> 00:01:32,540 To see if it can. 36 00:01:32,540 --> 00:01:36,940 Despite tears and tempers in a changing world. 37 00:01:36,940 --> 00:01:38,360 Nasty beast that we're visiting, 38 00:01:38,360 --> 00:01:40,179 could easily put me out of a job. 39 00:01:40,179 --> 00:01:42,041 ♪ And that's the point, yeah ♪ 40 00:01:42,041 --> 00:01:43,720 We can use computers to do 41 00:01:43,720 --> 00:01:46,260 crazy, exciting, creative things. 42 00:01:46,260 --> 00:01:49,037 The Experiment makes only one demand. 43 00:01:54,607 --> 00:01:56,857 (applause) 44 00:02:00,069 --> 00:02:02,649 Here we go, five, six, seven, and. 45 00:02:02,649 --> 00:02:05,875 ♪ Doe, a deer, a female deer ♪ 46 00:02:05,875 --> 00:02:07,888 ♪ Ray, a drop of golden sun ♪ 47 00:02:07,888 --> 00:02:09,760 Let's see your arms! 48 00:02:09,760 --> 00:02:11,840 Musical theater is big business, 49 00:02:11,840 --> 00:02:14,930 with millions of ardent fans worldwide. 50 00:02:14,930 --> 00:02:17,720 It's made a lot of people very happy. 51 00:02:17,720 --> 00:02:20,400 And some, very rich. 52 00:02:20,400 --> 00:02:22,583 I really, really like musicals. 53 00:02:22,583 --> 00:02:25,170 ♪ I'm in love, ooh ♪ 54 00:02:25,170 --> 00:02:28,553 But despite the love, the stats are sobering. 55 00:02:29,770 --> 00:02:32,223 Just one show in ten makes any money. 56 00:02:33,660 --> 00:02:36,123 And two out of ten lose everything. 57 00:02:37,600 --> 00:02:39,770 People know very well what they like 58 00:02:39,770 --> 00:02:42,520 and even talk about a good formula, 59 00:02:42,520 --> 00:02:44,010 but no one has ever really tried 60 00:02:44,010 --> 00:02:48,057 to crack it scientifically, until now. 61 00:02:48,057 --> 00:02:50,630 ("Together, Wherever We Go") 62 00:02:50,630 --> 00:02:51,952 North London. 63 00:02:51,952 --> 00:02:53,938 ♪ Wherever we go ♪ 64 00:02:53,938 --> 00:02:56,368 ♪ Whatever we do ♪ 65 00:02:56,368 --> 00:03:00,640 ♪ We're gonna go through it together ♪ 66 00:03:00,640 --> 00:03:03,230 Benjamin Till and Nathan Taylor 67 00:03:03,230 --> 00:03:05,510 are not scientists. 68 00:03:05,510 --> 00:03:08,750 Rather, they're an award-winning writer-composer duo, 69 00:03:08,750 --> 00:03:11,200 so passionate about musicals that they even 70 00:03:11,200 --> 00:03:13,043 turned their own wedding into one. 71 00:03:13,941 --> 00:03:18,000 ♪ How like the gays to marry in a musical ♪ 72 00:03:18,000 --> 00:03:19,060 For the Experiment, 73 00:03:19,060 --> 00:03:22,810 their task will be to curate the computer's creations 74 00:03:22,810 --> 00:03:25,100 into a script and musical score 75 00:03:25,100 --> 00:03:27,123 ready for humans to perform. 76 00:03:28,370 --> 00:03:30,330 My thing is, you know, I'm a composer, 77 00:03:30,330 --> 00:03:33,690 I write entirely from the perspective of emotion, 78 00:03:33,690 --> 00:03:35,240 and yes there's a little bit of technique 79 00:03:35,240 --> 00:03:37,780 in there as well, but I genuinely don't think 80 00:03:37,780 --> 00:03:40,620 you can teach a computer to feel. 81 00:03:40,620 --> 00:03:42,660 But maybe it's not about teaching a computer to feel. 82 00:03:42,660 --> 00:03:44,300 Maybe it's about teaching a computer 83 00:03:44,300 --> 00:03:46,453 what makes humans feel stuff. 84 00:03:49,110 --> 00:03:53,170 The Experiment may now begin. 85 00:03:53,170 --> 00:03:56,693 First, like with all experiments, we need some data. 86 00:03:58,250 --> 00:04:02,470 So our quest to make musical history begins where else, 87 00:04:02,470 --> 00:04:04,370 the Centre for Mathematical Sciences 88 00:04:04,370 --> 00:04:05,893 at Cambridge University. 89 00:04:07,024 --> 00:04:10,700 ♪ So prepare for the chance of a lifetime ♪ 90 00:04:10,700 --> 00:04:14,734 ♪ Be prepared for sensational news ♪ 91 00:04:14,734 --> 00:04:16,586 ♪ A shining new era ♪ 92 00:04:16,586 --> 00:04:18,576 ♪ Is tip-toeing nearer ♪ 93 00:04:18,576 --> 00:04:20,285 ♪ And where do we feature ♪ 94 00:04:20,285 --> 00:04:22,340 ♪ Just listen to teacher ♪ 95 00:04:22,340 --> 00:04:23,990 When this idea was first suggested, 96 00:04:23,990 --> 00:04:26,190 I thought it was bonkers, really. 97 00:04:26,190 --> 00:04:28,730 A musical's sort of soft and fluffy things, 98 00:04:28,730 --> 00:04:30,910 and I'm interested in hard science. 99 00:04:30,910 --> 00:04:32,920 But then, I realized well, you can measure 100 00:04:32,920 --> 00:04:36,110 the success of a musical, and you can measure 101 00:04:36,110 --> 00:04:37,807 various factors about the structure 102 00:04:37,807 --> 00:04:39,820 and the content of a musical, 103 00:04:39,820 --> 00:04:42,870 so why not then use our statistical techniques 104 00:04:42,870 --> 00:04:45,000 that enable us to predict outcomes 105 00:04:45,000 --> 00:04:48,704 to try to predict the success of a musical? 106 00:04:48,704 --> 00:04:50,060 ♪ One, two, three, four ♪ 107 00:04:50,060 --> 00:04:52,150 Professor Sir David Spiegelhalter 108 00:04:52,150 --> 00:04:54,560 and Dr. James Lloyd are experts 109 00:04:54,560 --> 00:04:57,183 in statistics and machine learning. 110 00:04:58,300 --> 00:05:00,750 Gather enough data about musicals, 111 00:05:00,750 --> 00:05:03,010 and they might be able to advise what 112 00:05:03,010 --> 00:05:04,973 and what not to put in ours. 113 00:05:08,400 --> 00:05:13,400 Over months, the team harvested data on over 1,700 musicals. 114 00:05:15,680 --> 00:05:18,460 All of it was fed to their computer algorithms 115 00:05:18,460 --> 00:05:21,073 to see if anything might be predictive of success. 116 00:05:24,238 --> 00:05:25,220 There are so many different types 117 00:05:25,220 --> 00:05:27,250 of data that relate to musicals. 118 00:05:27,250 --> 00:05:30,030 You've got the big databases of all the musicals 119 00:05:30,030 --> 00:05:32,730 and how long they lasted and how much money they took. 120 00:05:32,730 --> 00:05:35,670 You've got databases of what the reviewers said 121 00:05:35,670 --> 00:05:37,930 and the synopses written about the musicals. 122 00:05:37,930 --> 00:05:40,370 And then, to get the emotional content, 123 00:05:40,370 --> 00:05:42,180 and how people respond to them, 124 00:05:42,180 --> 00:05:44,040 you go to collect your own data. 125 00:05:44,040 --> 00:05:46,310 We've done all that, and I think created a database 126 00:05:46,310 --> 00:05:48,690 of the kind that's never existed before. 127 00:05:48,690 --> 00:05:50,180 First, every musical 128 00:05:50,180 --> 00:05:52,310 was plotted on the same graph. 129 00:05:52,310 --> 00:05:55,050 Based on critical or commercial success, 130 00:05:55,050 --> 00:05:57,250 they're split into four groups. 131 00:05:57,250 --> 00:06:00,150 Analyze this data to find out if any factors 132 00:06:00,150 --> 00:06:03,460 differentiate the hits from the flops, 133 00:06:03,460 --> 00:06:07,400 and we can make the first decisions about the new show. 134 00:06:07,400 --> 00:06:11,540 Initially, the statistics confirm the blindingly obvious. 135 00:06:11,540 --> 00:06:14,290 90% of musicals contain a love story. 136 00:06:14,290 --> 00:06:15,750 Also according to our surveys, 137 00:06:15,750 --> 00:06:18,463 90% of the musicals are set in the past. 138 00:06:19,820 --> 00:06:21,963 65% of them have a happy ending. 139 00:06:22,823 --> 00:06:25,360 And some conclusions can be drawn. 140 00:06:25,360 --> 00:06:27,030 The thing that's most striking about this 141 00:06:27,030 --> 00:06:29,790 is that the hits and the crowd pleasers 142 00:06:29,790 --> 00:06:32,460 have greater proportion of happy endings 143 00:06:32,460 --> 00:06:35,020 than the critically acclaimed and the flops. 144 00:06:35,020 --> 00:06:36,110 Interesting. 145 00:06:36,110 --> 00:06:37,295 The flops. 146 00:06:37,295 --> 00:06:39,220 ♪ Five, six, seven, eight ♪ 147 00:06:39,220 --> 00:06:41,180 Hours pass, and it's clear 148 00:06:41,180 --> 00:06:43,130 analyzing the data won't deliver 149 00:06:43,130 --> 00:06:45,823 one magical formula for success. 150 00:06:45,823 --> 00:06:49,019 (upbeat big band music) 151 00:06:49,019 --> 00:06:50,680 I like that, 'cause it seems so unlikely, 152 00:06:50,680 --> 00:06:51,843 yet it's so clear. 153 00:06:57,600 --> 00:06:59,060 But it will tell us 154 00:06:59,060 --> 00:07:02,323 what's more or less predictive of landing a hit. 155 00:07:06,760 --> 00:07:10,673 100,000 pieces of data about musicals in. 156 00:07:11,854 --> 00:07:13,220 Analyzed. 157 00:07:13,220 --> 00:07:17,563 And the Experiment now has a suggested recipe for success. 158 00:07:19,800 --> 00:07:22,180 Our musical should be a love story, 159 00:07:22,180 --> 00:07:24,830 feature a key death, be a quest 160 00:07:24,830 --> 00:07:28,730 to overcome loss, and it must dance. 161 00:07:28,730 --> 00:07:30,550 Multiple styles of music are better 162 00:07:30,550 --> 00:07:33,160 than it all being rather too samey. 163 00:07:33,160 --> 00:07:35,120 It needs to be set in the past, 164 00:07:35,120 --> 00:07:38,100 and the 1980s look most promising. 165 00:07:38,100 --> 00:07:40,630 We should avoid setting it in the USA. 166 00:07:40,630 --> 00:07:44,973 Our lead will be female and our ending should be happy. 167 00:07:46,750 --> 00:07:48,863 The list of ingredients goes on. 168 00:07:55,130 --> 00:07:59,010 But clever mathematics will only get us so far. 169 00:07:59,010 --> 00:08:00,500 Next, we must move from 170 00:08:00,500 --> 00:08:04,253 computer analysis to computer creativity. 171 00:08:06,472 --> 00:08:09,455 ♪ I wanna be a producer ♪ 172 00:08:09,455 --> 00:08:11,490 ♪ With a hit show on Broadway ♪ 173 00:08:11,490 --> 00:08:12,680 Top theater producer 174 00:08:12,680 --> 00:08:14,981 Neil Laidlaw has joined the team. 175 00:08:14,981 --> 00:08:16,141 ♪ I wanna be a producer ♪ 176 00:08:16,141 --> 00:08:18,630 ♪ Lunch at Sardi's every day ♪ 177 00:08:18,630 --> 00:08:19,930 He's secured just 178 00:08:19,930 --> 00:08:22,440 the West End venue to stage our show. 179 00:08:22,440 --> 00:08:25,040 The Arts Theatre, famous for its history 180 00:08:25,040 --> 00:08:27,367 of radical, experimental works. 181 00:08:27,367 --> 00:08:29,977 ♪ I wanna be a producer ♪ 182 00:08:29,977 --> 00:08:32,970 ♪ And drive those chorus girls insane ♪ 183 00:08:32,970 --> 00:08:35,020 What was it like for you hearing 184 00:08:35,020 --> 00:08:37,270 the rather quirky aspect of what we're doing? 185 00:08:38,410 --> 00:08:40,659 Well, I turned it down a number of times. 186 00:08:40,659 --> 00:08:43,030 (laughing) 187 00:08:43,030 --> 00:08:47,840 Mainly because I think it's the real unknown factor, 188 00:08:47,840 --> 00:08:50,223 but when it came to it and I actually thought, 189 00:08:51,450 --> 00:08:53,500 in 20 years time, am I gonna look back 190 00:08:53,500 --> 00:08:56,120 and kick myself and go, why didn't I try it. 191 00:08:56,120 --> 00:08:57,900 'Cause the bottom line is, I mean, 192 00:08:57,900 --> 00:08:59,920 in some ways, it's completely liberating, 193 00:08:59,920 --> 00:09:03,610 because if the show's ghastly, it's not really my fault. 194 00:09:03,610 --> 00:09:07,810 And really what drew me in was the thought of, 195 00:09:07,810 --> 00:09:09,587 but what could come out of it. 196 00:09:12,521 --> 00:09:14,323 ♪ God, I hope I get it ♪ 197 00:09:14,323 --> 00:09:16,200 ♪ I hope I get it ♪ 198 00:09:16,200 --> 00:09:18,680 Hot on Neil's heels, the Experiment 199 00:09:18,680 --> 00:09:22,273 has director Luke Sheppard in its sights. 200 00:09:24,520 --> 00:09:26,540 The hope is that he, too, will be hooked 201 00:09:26,540 --> 00:09:30,760 by the exciting unknowns of a computer generated show, 202 00:09:30,760 --> 00:09:33,305 about which he currently knows nothing. 203 00:09:33,305 --> 00:09:35,147 ♪ How many people does he ♪ 204 00:09:35,147 --> 00:09:39,130 The key point is to come up with a musical 205 00:09:39,130 --> 00:09:41,283 that is entirely written by computers. 206 00:09:42,850 --> 00:09:43,810 Wow. 207 00:09:43,810 --> 00:09:45,300 Okay. 208 00:09:45,300 --> 00:09:46,133 Interesting. 209 00:09:48,090 --> 00:09:49,131 When do we start? 210 00:09:49,131 --> 00:09:51,230 (all laughing) 211 00:09:51,230 --> 00:09:52,577 We don't expect that a computer, 212 00:09:52,577 --> 00:09:54,190 you can push a button and out pops 213 00:09:54,190 --> 00:09:56,230 West Side Story or My Fair Lady. 214 00:09:56,230 --> 00:09:59,190 None of us know at this point how much of this is possible. 215 00:09:59,190 --> 00:10:02,130 Is that something that causes you any problems? 216 00:10:02,130 --> 00:10:03,590 It doesn't cause me any problems. 217 00:10:03,590 --> 00:10:07,030 In fact, it draws together I think 218 00:10:07,030 --> 00:10:09,070 a lot of the different areas that I tend to work in, 219 00:10:09,070 --> 00:10:11,110 and I think the thrill of it potentially being 220 00:10:11,110 --> 00:10:14,540 a failure is probably what makes it worth doing. 221 00:10:14,540 --> 00:10:16,260 It'd be interesting to find out 222 00:10:16,260 --> 00:10:19,150 what the kind of computer's voice is within that, 223 00:10:19,150 --> 00:10:21,163 and how you engage in that dialogue. 224 00:10:23,160 --> 00:10:25,830 One thing that we do think about a lot 225 00:10:25,830 --> 00:10:29,536 is whether computers can be an actual creative partner. 226 00:10:29,536 --> 00:10:31,360 Are they worth collaborating with? 227 00:10:31,360 --> 00:10:32,650 Are they worth working with? 228 00:10:32,650 --> 00:10:34,580 The computer programs are essentially, 229 00:10:34,580 --> 00:10:35,413 they're in their infancy. 230 00:10:35,413 --> 00:10:37,960 They're students and we are learning 231 00:10:37,960 --> 00:10:40,230 how to make them more creative. 232 00:10:40,230 --> 00:10:44,570 We are learning how to turn things around 233 00:10:44,570 --> 00:10:46,410 from making computers creative 234 00:10:46,410 --> 00:10:49,330 to letting computers develop their own creativity, 235 00:10:49,330 --> 00:10:50,886 as if they were students. 236 00:10:50,886 --> 00:10:54,650 ♪ Let's start at the very beginning ♪ 237 00:10:54,650 --> 00:10:56,850 For the Experiment, we must now see 238 00:10:56,850 --> 00:10:59,860 where the computers can get creative. 239 00:10:59,860 --> 00:11:02,132 And just as humans do, it all starts 240 00:11:02,132 --> 00:11:04,854 with the spark of an idea. 241 00:11:04,854 --> 00:11:06,373 ♪ Do, re, mi ♪ 242 00:11:06,373 --> 00:11:08,389 ♪ Do, re, mi ♪ 243 00:11:08,389 --> 00:11:09,510 ♪ Do, re, mi ♪ 244 00:11:09,510 --> 00:11:11,260 This team of academics 245 00:11:11,260 --> 00:11:13,670 from London's Goldsmiths University 246 00:11:13,670 --> 00:11:15,830 have agreed to introduce their very own 247 00:11:15,830 --> 00:11:19,733 idea generating computer to the world of musicals. 248 00:11:21,282 --> 00:11:22,780 It's our job to get software 249 00:11:22,780 --> 00:11:26,060 to invent ideas, fictional ideas. 250 00:11:26,060 --> 00:11:28,210 And so we are building this machine 251 00:11:28,210 --> 00:11:32,080 called the What-If Machine, and it's designed 252 00:11:32,080 --> 00:11:35,600 to invent fictions which possibly could be used 253 00:11:35,600 --> 00:11:37,950 as the heart of cultural artifacts, 254 00:11:37,950 --> 00:11:40,620 like short stories, poems, jokes, 255 00:11:40,620 --> 00:11:42,763 and of course, musical theater pieces. 256 00:11:45,330 --> 00:11:48,040 The What-If Machine offers endless ideas 257 00:11:48,040 --> 00:11:52,490 for different kinds of story or character. 258 00:11:52,490 --> 00:11:56,273 Its suggestions range from the brilliant to the barmy. 259 00:11:57,608 --> 00:12:02,060 Spinning. (laughs) 260 00:12:02,060 --> 00:12:04,980 What if a lordly brat was supposed to annoy 261 00:12:04,980 --> 00:12:06,650 a sibling but couldn't because of 262 00:12:06,650 --> 00:12:09,660 the embarrassing woman he had fallen in love with? 263 00:12:09,660 --> 00:12:13,990 I mean, that is the most complicated, bizarre plot. 264 00:12:13,990 --> 00:12:15,943 A lordly brat. Go, let's do it. 265 00:12:16,960 --> 00:12:19,700 The computer relies on masses of data 266 00:12:19,700 --> 00:12:23,290 about how stuff interacts in the real world. 267 00:12:23,290 --> 00:12:25,490 Show me another love one, because we've been told 268 00:12:25,490 --> 00:12:28,033 that there will be love in this. 269 00:12:28,999 --> 00:12:29,832 Did the Cambridge team 270 00:12:29,832 --> 00:12:31,800 tell you that, good. They told us that love was. 271 00:12:31,800 --> 00:12:33,490 600 of the computer's 272 00:12:33,490 --> 00:12:36,293 brand new ideas are ready for perusal. 273 00:12:37,220 --> 00:12:39,210 Okay, Ben, find us our hit. 274 00:12:39,210 --> 00:12:40,710 (laughing) 275 00:12:40,710 --> 00:12:43,120 What if there was a poor boy who was born with a hop, 276 00:12:43,120 --> 00:12:44,750 which made him very good at jumping, 277 00:12:44,750 --> 00:12:47,053 thanks to which he became a very famous miner? 278 00:12:48,430 --> 00:12:50,520 The computer analysis has already said 279 00:12:50,520 --> 00:12:54,300 that ours will be a quest led by a female character, 280 00:12:54,300 --> 00:12:56,793 involving a love story and death. 281 00:12:57,660 --> 00:13:00,920 What we've sort of learned is that 282 00:13:00,920 --> 00:13:03,400 the stakes need to be really high. 283 00:13:03,400 --> 00:13:05,640 What if there was a wounded soldier 284 00:13:05,640 --> 00:13:08,590 who had to learn how to understand a child 285 00:13:08,590 --> 00:13:10,157 in order to find true love? 286 00:13:10,157 --> 00:13:11,563 I like that one. Yeah, it's good. 287 00:13:12,690 --> 00:13:13,523 There's some back here. 288 00:13:13,523 --> 00:13:14,860 Over there? Yeah. 289 00:13:14,860 --> 00:13:15,970 I found a great one. 290 00:13:15,970 --> 00:13:19,300 What if there was a doctor who found a magical medicine 291 00:13:19,300 --> 00:13:21,623 that would allow him to heal disharmony? 292 00:13:23,240 --> 00:13:26,330 After quite some time treading the boards, 293 00:13:26,330 --> 00:13:29,570 we have a shortlist of computer generated ideas 294 00:13:29,570 --> 00:13:31,810 for our new musical. 295 00:13:31,810 --> 00:13:33,697 I'm holding the best one in my hand, so. 296 00:13:33,697 --> 00:13:35,430 I don't know, I may have it here. 297 00:13:35,430 --> 00:13:37,060 But our experiment to create 298 00:13:37,060 --> 00:13:41,210 a computer generated hit musical has come to Spain, 299 00:13:41,210 --> 00:13:43,373 and is about to get a whole lot harder. 300 00:13:44,220 --> 00:13:47,123 As with all great art, we need to tell a story. 301 00:13:48,682 --> 00:13:51,432 (flamenco music) 302 00:13:52,940 --> 00:13:55,770 Story telling is one of the most complex activities 303 00:13:55,770 --> 00:13:57,780 that the human brain can engage on, 304 00:13:57,780 --> 00:14:01,270 because it draws upon knowledge about the world, 305 00:14:01,270 --> 00:14:03,950 knowledge about people, how people react, 306 00:14:03,950 --> 00:14:06,400 about character, about personality, 307 00:14:06,400 --> 00:14:09,600 about social norm, about emotions. 308 00:14:09,600 --> 00:14:11,260 And this is very, very difficult. 309 00:14:11,260 --> 00:14:14,010 Computer can only do it if we provide it 310 00:14:14,010 --> 00:14:17,777 with enough information to know about these things. 311 00:14:19,580 --> 00:14:22,180 Dr. Pablo Gervas is a world expert 312 00:14:22,180 --> 00:14:24,550 in understanding the basic building blocks 313 00:14:24,550 --> 00:14:26,223 that stories are made of. 314 00:14:27,210 --> 00:14:30,710 Pablo's computer system, which he's called PropperWryter, 315 00:14:30,710 --> 00:14:33,780 is fed a set of hit musical plots. 316 00:14:33,780 --> 00:14:35,560 It works out if there are patterns 317 00:14:35,560 --> 00:14:38,190 in how their stories are built, 318 00:14:38,190 --> 00:14:41,607 and then starts to build plots entirely of its own. 319 00:14:44,150 --> 00:14:48,610 We've decided to start with an abstract view of the story, 320 00:14:48,610 --> 00:14:52,020 and try and find a core view 321 00:14:52,020 --> 00:14:54,030 of what is happening in the story. 322 00:14:54,030 --> 00:14:56,690 Things like someone falls in love, 323 00:14:56,690 --> 00:14:59,920 or two people struggle against 324 00:14:59,920 --> 00:15:01,940 one another, that sort of thing. 325 00:15:01,940 --> 00:15:04,520 We've asked a whole lot of volunteers 326 00:15:04,520 --> 00:15:08,520 to look at musicals and tell us 327 00:15:08,520 --> 00:15:10,230 how they see these musicals in terms of 328 00:15:10,230 --> 00:15:11,550 these big things happening. 329 00:15:11,550 --> 00:15:13,920 With this data, Pablo can track 330 00:15:13,920 --> 00:15:17,310 love, danger, or happiness developing 331 00:15:17,310 --> 00:15:20,180 as any musical plot unfolds. 332 00:15:20,180 --> 00:15:24,550 This is South Pacific in terms of these plot elements. 333 00:15:24,550 --> 00:15:25,790 You can see the plot line there, 334 00:15:25,790 --> 00:15:29,130 and how love evolves or hate, there's a spike of hate-- 335 00:15:29,130 --> 00:15:30,728 Why would the spike of hate be there? 336 00:15:30,728 --> 00:15:31,630 Well that'd be the end of act one 337 00:15:31,630 --> 00:15:34,320 where Nellie realizes that he was married 338 00:15:34,320 --> 00:15:35,900 to a Polynesian lady before. 339 00:15:35,900 --> 00:15:38,320 Yeah, the racist bit. Ah, the racist bit. 340 00:15:38,320 --> 00:15:40,360 It's labeled as class differences. 341 00:15:40,360 --> 00:15:41,200 Right, there we are, then. 342 00:15:41,200 --> 00:15:44,750 Click here and you get what the system's put together, 343 00:15:44,750 --> 00:15:46,320 which is basically a sequence. 344 00:15:46,320 --> 00:15:48,210 So it's adding its own information. 345 00:15:48,210 --> 00:15:51,220 It's trying to make a coherent plot, 346 00:15:51,220 --> 00:15:52,620 out of what you've given it. 347 00:15:53,990 --> 00:15:57,730 Click again, it will look for another combination. 348 00:15:57,730 --> 00:16:00,090 In each case it's building a new plot. 349 00:16:00,090 --> 00:16:04,670 The most important part of getting a musical right, 350 00:16:04,670 --> 00:16:06,580 for both of us, I think I'm right in saying, 351 00:16:06,580 --> 00:16:11,170 is getting the story structure underpinning everything else. 352 00:16:11,170 --> 00:16:13,950 Despite never having been tested before, 353 00:16:13,950 --> 00:16:16,320 Pablo's plot-generating computer 354 00:16:16,320 --> 00:16:19,233 will deliver the backbone of our new show. 355 00:16:20,197 --> 00:16:22,447 ("9 to 5") 356 00:16:24,930 --> 00:16:27,950 Back in London, at Human HQ, 357 00:16:27,950 --> 00:16:29,930 it's finally time for Ben and Nathan 358 00:16:29,930 --> 00:16:31,833 to make some big decisions. 359 00:16:32,770 --> 00:16:35,160 First, we need our basic idea. 360 00:16:35,160 --> 00:16:37,920 Which what-if will they go for? 361 00:16:37,920 --> 00:16:38,753 What if there was 362 00:16:38,753 --> 00:16:40,220 a wounded soldier who had to learn 363 00:16:40,220 --> 00:16:43,083 how to understand a child in order to find true love? 364 00:16:44,431 --> 00:16:47,020 It's got such romance in it, 365 00:16:47,020 --> 00:16:48,070 and it's. Totally. 366 00:16:49,405 --> 00:16:51,893 Our musical has its big idea. 367 00:16:54,510 --> 00:16:56,510 The Experiment's method for deciding 368 00:16:56,510 --> 00:16:58,260 the real-world inspiration for 369 00:16:58,260 --> 00:17:00,620 the musical is perfectly simple. 370 00:17:00,620 --> 00:17:03,020 Put all the keywords from the computer's idea 371 00:17:03,020 --> 00:17:07,270 into a search engine, along with possible settings, 372 00:17:07,270 --> 00:17:11,442 and let the algorithm suggest where to find our female lead. 373 00:17:11,442 --> 00:17:12,275 Alright. 374 00:17:12,275 --> 00:17:13,170 Did you mean (laughs). 375 00:17:15,400 --> 00:17:18,987 When Greenham Common Women's Peace Camp Songbook comes up. 376 00:17:18,987 --> 00:17:21,240 No, you've got to write your own. 377 00:17:21,240 --> 00:17:22,843 No, I know, but. 378 00:17:22,843 --> 00:17:25,863 ♪ We don't want your cruise ♪ 379 00:17:25,863 --> 00:17:28,886 ♪ We have life to lose ♪ 380 00:17:28,886 --> 00:17:32,003 ♪ It's not too late to choose ♪ 381 00:17:32,003 --> 00:17:35,170 ♪ Jump down, Greenham ♪ 382 00:17:36,623 --> 00:17:38,158 Yay. 383 00:17:38,158 --> 00:17:39,104 Happy? 384 00:17:39,104 --> 00:17:41,087 Yeah, I'm really happy. (laughs) 385 00:17:41,087 --> 00:17:42,537 Are you gonna cry? 386 00:17:45,860 --> 00:17:46,910 From the data, 387 00:17:46,910 --> 00:17:49,783 our musical has its unexpected setting. 388 00:17:51,880 --> 00:17:53,910 Greenham Common Air Base was the site of 389 00:17:53,910 --> 00:17:56,853 the largest women's peace camp in British history. 390 00:17:58,490 --> 00:18:01,490 Established in 1981, the protest aimed 391 00:18:01,490 --> 00:18:03,140 to prevent the US Air Force 392 00:18:03,140 --> 00:18:05,753 sighting nuclear cruise missiles there. 393 00:18:08,420 --> 00:18:11,470 Quite a lot of my childhood was spent on a commune 394 00:18:11,470 --> 00:18:15,610 in Bedfordshire, which was run by CND women. 395 00:18:15,610 --> 00:18:20,110 The idea that a computer can lead me home 396 00:18:20,110 --> 00:18:22,410 to something that is of deep significance, 397 00:18:22,410 --> 00:18:27,123 personal significance, is actually quite extraordinary. 398 00:18:28,394 --> 00:18:32,023 ♪ At the end of the day you're another day older ♪ 399 00:18:32,023 --> 00:18:34,440 With a basic idea and a setting, 400 00:18:34,440 --> 00:18:37,790 it's time to use PropperWryter, Pablo's machine, 401 00:18:37,790 --> 00:18:40,813 to generate the best possible story. 402 00:18:40,813 --> 00:18:43,115 ♪ What is it for ♪ 403 00:18:43,115 --> 00:18:44,623 ♪ One day less to be living ♪ 404 00:18:44,623 --> 00:18:46,158 From the beginning, we're doing-- 405 00:18:46,158 --> 00:18:48,090 ♪ At the end of the day you're another day colder ♪ 406 00:18:48,090 --> 00:18:50,570 Then, maybe we have a plot. 407 00:18:50,570 --> 00:18:53,430 So on the wall now, we have a series of timelines, 408 00:18:53,430 --> 00:18:55,940 which we've been creating over the last couple of weeks. 409 00:18:55,940 --> 00:18:58,350 We have the Greenham Common timeline, 410 00:18:58,350 --> 00:19:00,230 which is at the top, which tells us 411 00:19:00,230 --> 00:19:02,350 what happened in what month, what year. 412 00:19:02,350 --> 00:19:05,070 And then we have the Cambridge timeline, 413 00:19:05,070 --> 00:19:07,050 which tells us what we need 414 00:19:07,050 --> 00:19:09,650 to achieve overall in our musical. 415 00:19:09,650 --> 00:19:13,880 And then we have the idea of this soldier and the child, 416 00:19:13,880 --> 00:19:18,880 which has now been fleshed out in pink by Pablo's machine. 417 00:19:19,500 --> 00:19:21,590 So this basically, this is pure. 418 00:19:21,590 --> 00:19:23,940 This is all computers. 419 00:19:23,940 --> 00:19:25,870 We now have to try and work out 420 00:19:25,870 --> 00:19:28,833 what does this dictate for us. 421 00:19:28,833 --> 00:19:31,416 (upbeat music) 422 00:19:32,730 --> 00:19:33,770 It's time to share 423 00:19:33,770 --> 00:19:36,460 the exciting news with Neil and Luke, 424 00:19:36,460 --> 00:19:38,493 our show's producer and director. 425 00:19:40,430 --> 00:19:42,460 The minute I saw Greenham, I thought, 426 00:19:42,460 --> 00:19:44,913 why's no one set a musical there before? 427 00:19:45,830 --> 00:19:47,980 The computers have worked hard, 428 00:19:47,980 --> 00:19:50,740 but it's the human who've had to make the choices 429 00:19:50,740 --> 00:19:53,243 and they've a huge task ahead. 430 00:19:54,910 --> 00:19:57,960 Because the individual components of software 431 00:19:57,960 --> 00:20:00,230 that have been used in this project 432 00:20:00,230 --> 00:20:02,730 don't have an awareness about the bigger picture, 433 00:20:02,730 --> 00:20:04,170 don't have an awareness about 434 00:20:04,170 --> 00:20:06,620 what each of the other things are doing, 435 00:20:06,620 --> 00:20:08,740 you need humans in the loop to say, 436 00:20:08,740 --> 00:20:10,931 actually, that's the storyline I want. 437 00:20:10,931 --> 00:20:12,680 That's the chord sequence I want. 438 00:20:12,680 --> 00:20:13,780 That's the lyric I want. 439 00:20:13,780 --> 00:20:15,980 Because they all fit together. 440 00:20:15,980 --> 00:20:18,080 There's a lot there for us to go and get our teeth into. 441 00:20:18,080 --> 00:20:20,160 There's also a lot for us to work through. 442 00:20:20,160 --> 00:20:21,400 It's so exciting. 443 00:20:21,400 --> 00:20:24,400 Every decision we make needs to come back 444 00:20:24,400 --> 00:20:27,890 to the sets of data that got us there in the first place. 445 00:20:27,890 --> 00:20:29,210 And the music? 446 00:20:29,210 --> 00:20:30,043 Who knows. 447 00:20:30,043 --> 00:20:32,130 There'll be as much of it as we can possibly provide. 448 00:20:32,130 --> 00:20:34,166 Music to my ears, just without the music. 449 00:20:34,166 --> 00:20:35,350 (laughing) 450 00:20:35,350 --> 00:20:36,440 It's coming. 451 00:20:36,440 --> 00:20:37,930 ♪ Write what you know, so they say ♪ 452 00:20:37,930 --> 00:20:39,963 ♪ All I know is I don't know what to write ♪ 453 00:20:39,963 --> 00:20:41,486 ♪ Or the right way to write it ♪ 454 00:20:41,486 --> 00:20:43,342 ♪ This is big, lady, don't screw it up ♪ 455 00:20:43,342 --> 00:20:47,220 ♪ This is not some little vaudeville I'm reviewing ♪ 456 00:20:47,220 --> 00:20:48,600 For the Experiment, 457 00:20:48,600 --> 00:20:51,370 with what feels like a winning show in hand, 458 00:20:51,370 --> 00:20:53,060 the next challenge, this being 459 00:20:53,060 --> 00:20:55,840 a musical after all, are the songs. 460 00:20:55,840 --> 00:20:58,243 Specifically, their lyrics. 461 00:20:58,243 --> 00:20:59,371 ♪ 'Cause as I may have mentioned ♪ 462 00:20:59,371 --> 00:21:02,830 ♪ I have no clue what I'm doing ♪ 463 00:21:02,830 --> 00:21:05,681 Dr. James Lloyd and Dr. Alex Davies 464 00:21:05,681 --> 00:21:07,300 are part of the Cambridge team 465 00:21:07,300 --> 00:21:10,350 that generated our recipe for success. 466 00:21:10,350 --> 00:21:12,690 Being quite extraordinarily smart, 467 00:21:12,690 --> 00:21:15,460 they'll be providing a new computer system 468 00:21:15,460 --> 00:21:19,190 to which Ben and Nathan will soon be shackled. 469 00:21:19,190 --> 00:21:20,680 We were really interested in looking into 470 00:21:20,680 --> 00:21:23,230 some advances in neural network techniques 471 00:21:23,230 --> 00:21:25,130 that have happened recently. 472 00:21:25,130 --> 00:21:27,650 In understanding language and to get a bit more 473 00:21:27,650 --> 00:21:28,900 of a handle on this, we decided 474 00:21:28,900 --> 00:21:32,790 to try and teach the computer to write poetry. 475 00:21:32,790 --> 00:21:33,886 And so we just-- 476 00:21:33,886 --> 00:21:35,610 Just like that. Just like that. 477 00:21:35,610 --> 00:21:37,900 The reason that we call it a neural network 478 00:21:37,900 --> 00:21:40,652 is because its architecture, or the way it's built, 479 00:21:40,652 --> 00:21:44,610 is based on the low-level architecture of a brain. 480 00:21:44,610 --> 00:21:48,300 This computer currently has poetry on its mind. 481 00:21:48,300 --> 00:21:50,700 This is after reading 200 poems, 482 00:21:50,700 --> 00:21:53,110 we said, please try and generate some poetry, 483 00:21:53,110 --> 00:21:53,960 and it does this. 484 00:21:56,380 --> 00:21:58,417 That screen actually makes me feel sick. 485 00:21:58,417 --> 00:21:59,740 (laughing) 486 00:21:59,740 --> 00:22:01,810 So it hasn't done much, but it's learned 487 00:22:01,810 --> 00:22:04,130 some common letters, you can see there are lots of, 488 00:22:04,130 --> 00:22:07,470 quite a few vowels, Ts, Ss, Hs, that sort of thing. 489 00:22:07,470 --> 00:22:10,440 10,000 poems later. 490 00:22:10,440 --> 00:22:12,770 Searching the forest of blood. 491 00:22:12,770 --> 00:22:15,830 All stars tattered through a clock from the moon. 492 00:22:15,830 --> 00:22:17,918 My god! 493 00:22:17,918 --> 00:22:19,269 That's fabulous. 494 00:22:19,269 --> 00:22:20,102 Brilliant. 495 00:22:20,102 --> 00:22:23,990 I'd be really interested to see if one could 496 00:22:23,990 --> 00:22:26,870 put the entire canon of musical theater lyrics into it 497 00:22:26,870 --> 00:22:30,913 and see what it starts to write from there. 498 00:22:32,090 --> 00:22:33,550 Shall we see what it does? 499 00:22:33,550 --> 00:22:34,383 Have you done it? 500 00:22:34,383 --> 00:22:35,522 Yep. Amazing. 501 00:22:35,522 --> 00:22:36,520 (laughing) 502 00:22:36,520 --> 00:22:39,269 A girl is a carnival star that's been born. 503 00:22:39,269 --> 00:22:40,870 (laughing) 504 00:22:40,870 --> 00:22:42,380 That's brilliant. 505 00:22:42,380 --> 00:22:44,800 And the figure is drunk, and more. 506 00:22:44,800 --> 00:22:47,800 And the world is dead and so-so. 507 00:22:47,800 --> 00:22:52,010 The friends are long ago and the stars were blown away. 508 00:22:52,010 --> 00:22:53,710 Wow. That is brilliant. 509 00:22:53,710 --> 00:22:55,690 Our new lyricist in training 510 00:22:55,690 --> 00:22:58,180 can only be of so much use without 511 00:22:58,180 --> 00:23:01,903 the last part of the Experiment, music. 512 00:23:03,898 --> 00:23:07,640 ♪ Not a day goes by ♪ 513 00:23:07,640 --> 00:23:10,460 We begin with Dr. Bob Sturm. 514 00:23:10,460 --> 00:23:12,975 He's a specialist in machine listening. 515 00:23:12,975 --> 00:23:16,366 ♪ When does it end ♪ 516 00:23:16,366 --> 00:23:18,880 To get a computer to listen to music as humans do, 517 00:23:18,880 --> 00:23:21,030 is an extremely difficult challenge. 518 00:23:21,030 --> 00:23:23,950 The way our ears work, we want to mimic that 519 00:23:23,950 --> 00:23:27,850 as much as we can so we get an idea of what's playing. 520 00:23:27,850 --> 00:23:29,170 Is it a musical note? 521 00:23:29,170 --> 00:23:31,560 And we build these conceptions higher and higher 522 00:23:31,560 --> 00:23:34,830 to reach chords and eventually melodies 523 00:23:34,830 --> 00:23:37,500 and styles or genres. 524 00:23:37,500 --> 00:23:39,683 It's a very complex process. 525 00:23:42,270 --> 00:23:43,450 Bob is here because 526 00:23:43,450 --> 00:23:46,240 once computers can listen to musicals, 527 00:23:46,240 --> 00:23:48,643 then they can start to learn how to write them. 528 00:23:50,110 --> 00:23:52,040 So we have Abby's recording here, 529 00:23:52,040 --> 00:23:53,410 and what I'm going to show you 530 00:23:53,410 --> 00:23:57,190 is how we make a machine listen to the music 531 00:23:57,190 --> 00:23:59,640 and pick apart what we're interested in. 532 00:23:59,640 --> 00:24:00,840 These are frequencies, but let's 533 00:24:00,840 --> 00:24:02,350 make this more musically meaningful. 534 00:24:02,350 --> 00:24:04,340 So we can break all of these frequencies down 535 00:24:04,340 --> 00:24:06,530 into the 12 notes of the scale. 536 00:24:06,530 --> 00:24:09,290 We can now perform a chord analysis, 537 00:24:09,290 --> 00:24:10,680 and it's saying, oh, there's a lot 538 00:24:10,680 --> 00:24:13,370 of notes in here in common with D major. 539 00:24:13,370 --> 00:24:14,203 It is, yeah. 540 00:24:14,203 --> 00:24:15,740 There's your A, there's your D, and there's your F sharp. 541 00:24:15,740 --> 00:24:17,360 Playing the computer's chords 542 00:24:17,360 --> 00:24:20,470 along to the original song shows 543 00:24:20,470 --> 00:24:22,853 the computer heard it right. 544 00:24:22,853 --> 00:24:26,114 There's a D, D, yeah. 545 00:24:26,114 --> 00:24:29,531 (Ben plays piano chords) 546 00:24:31,114 --> 00:24:31,947 Interesting. 547 00:24:31,947 --> 00:24:33,250 When it comes to music, 548 00:24:33,250 --> 00:24:37,733 Ben's relationship with computers is a little strained. 549 00:24:39,960 --> 00:24:43,113 I use computers as much as I need to use computers. 550 00:24:45,110 --> 00:24:47,400 As much as they can help me directly. 551 00:24:47,400 --> 00:24:50,610 So from the perspective of whether or not 552 00:24:50,610 --> 00:24:53,110 I think a computer is capable of writing music, 553 00:24:53,110 --> 00:24:56,373 well yes, I hope they are, for this project. 554 00:24:57,640 --> 00:25:00,090 The rest of me hopes they're not at all, 555 00:25:00,090 --> 00:25:02,710 for my own life, for my own career. 556 00:25:02,710 --> 00:25:05,330 Because this nasty beast that we're visiting 557 00:25:05,330 --> 00:25:07,258 could easily put me out of a job. 558 00:25:07,258 --> 00:25:09,641 ♪ Gonna buy me a ticket ♪ 559 00:25:09,641 --> 00:25:13,058 ♪ And ride on that B & O ♪ 560 00:25:14,160 --> 00:25:16,350 In search of computer music, 561 00:25:16,350 --> 00:25:19,853 there's no need for Silicon Valley when you have Durham. 562 00:25:22,760 --> 00:25:25,920 (dramatic piano music) 563 00:25:25,920 --> 00:25:28,570 Dr. Nick Collins is a leading researcher 564 00:25:28,570 --> 00:25:31,783 and composer of computer generated music. 565 00:25:49,280 --> 00:25:51,300 So I'm pretty sure what you were playing 566 00:25:51,300 --> 00:25:54,650 just then was written by a computer, 567 00:25:54,650 --> 00:25:57,610 but only because I'm sitting here 568 00:25:57,610 --> 00:26:00,000 with you now and you chose to play that. 569 00:26:00,000 --> 00:26:02,650 You're quite right that the pieces 570 00:26:02,650 --> 00:26:05,840 that I just played to you were created 571 00:26:05,840 --> 00:26:08,060 via the medium of a computer. 572 00:26:08,060 --> 00:26:10,550 How much do you know about musical theater? 573 00:26:10,550 --> 00:26:13,390 I've been to the odd musical theater show in my life, 574 00:26:13,390 --> 00:26:16,130 but I would by no means claim to be an expert, 575 00:26:16,130 --> 00:26:17,880 and I guess that's why you're here. 576 00:26:19,030 --> 00:26:22,843 What I have got is quite a large amount of data. 577 00:26:24,740 --> 00:26:26,100 Bob's team's machine 578 00:26:26,100 --> 00:26:29,450 listened to hundreds of musical theater songs. 579 00:26:29,450 --> 00:26:34,450 Nick used that data to teach his computer about musicals. 580 00:26:34,610 --> 00:26:37,423 And with that, it can begin to compose. 581 00:26:39,890 --> 00:26:42,277 This is its first try. 582 00:26:42,277 --> 00:26:44,150 And as you can see, I've just generated 583 00:26:44,150 --> 00:26:46,030 a show tune by my program, 584 00:26:46,030 --> 00:26:48,200 I call my program Android Lloyd Webber. 585 00:26:48,200 --> 00:26:50,880 That is a crackingly awful pun. 586 00:26:50,880 --> 00:26:52,030 Thank you. 587 00:26:52,030 --> 00:26:56,480 You can see that it's a bit stuck initially on one note. 588 00:26:56,480 --> 00:26:58,968 Although that is not necessarily a bad thing. 589 00:26:58,968 --> 00:26:59,801 Yeah? 590 00:26:59,801 --> 00:27:02,301 (piano music) 591 00:27:06,620 --> 00:27:08,406 So we have the particular bar here that goes. 592 00:27:08,406 --> 00:27:10,906 (piano music) 593 00:27:19,270 --> 00:27:20,700 What I'm not picking up from it 594 00:27:20,700 --> 00:27:23,363 is any recognizable chord progressions. 595 00:27:24,280 --> 00:27:27,330 Which is something that I think a lot of popular music, 596 00:27:27,330 --> 00:27:29,621 particularly musical theater, is based on. 597 00:27:29,621 --> 00:27:30,454 Yeah, yeah. 598 00:27:30,454 --> 00:27:32,300 Okay, so when I run this program, 599 00:27:32,300 --> 00:27:35,150 every time I run it we get back a totally different tune. 600 00:27:36,497 --> 00:27:38,997 (piano music) 601 00:27:50,310 --> 00:27:51,450 And so on. 602 00:27:51,450 --> 00:27:53,933 The sort of thing I was writing when I was about eight. 603 00:27:55,320 --> 00:27:56,580 Maybe? Maybe? 604 00:27:56,580 --> 00:28:01,393 How do you then turn that into something that sounds good? 605 00:28:03,130 --> 00:28:04,870 Ben and Nick spend the afternoon 606 00:28:04,870 --> 00:28:08,660 exploring Android's prolific compositional strengths 607 00:28:08,660 --> 00:28:10,760 and weaknesses before getting to 608 00:28:10,760 --> 00:28:12,743 a rather important question. 609 00:28:13,850 --> 00:28:18,100 What is it about computer writing music 610 00:28:18,100 --> 00:28:20,503 that you find so compelling? 611 00:28:21,870 --> 00:28:23,900 There's all sorts of reasons to do it. 612 00:28:23,900 --> 00:28:26,410 One is that you can make new music, 613 00:28:26,410 --> 00:28:28,130 and perhaps make new music in ways 614 00:28:28,130 --> 00:28:31,570 that you yourself wouldn't have thought of. 615 00:28:31,570 --> 00:28:35,100 If I was to die in the next five minutes, 616 00:28:35,100 --> 00:28:37,420 in a certain sense my programs would survive me. 617 00:28:37,420 --> 00:28:38,840 So there's a part of myself 618 00:28:38,840 --> 00:28:40,920 that's encapsulated within that code. 619 00:28:40,920 --> 00:28:44,320 One of the reasons why I think I do what I do 620 00:28:44,320 --> 00:28:47,300 is because it's about legacy 621 00:28:48,550 --> 00:28:52,790 and leaving the world something that I created, 622 00:28:52,790 --> 00:28:55,760 something that's mine that 623 00:28:55,760 --> 00:28:57,773 other people can enjoy, I suppose. 624 00:28:58,630 --> 00:29:01,360 But I can flood the world with pieces 625 00:29:01,360 --> 00:29:03,730 that I've composed forever, 626 00:29:03,730 --> 00:29:07,000 whereas you have a fixed base of pieces 627 00:29:07,000 --> 00:29:09,300 that will become diminishingly small 628 00:29:09,300 --> 00:29:12,290 with respect to all the future composers. 629 00:29:12,290 --> 00:29:15,023 Yes, but my soul exists in my music. 630 00:29:16,010 --> 00:29:17,880 But my soul would exist in my programs 631 00:29:17,880 --> 00:29:20,333 and through the program, into the further music. 632 00:29:21,510 --> 00:29:22,553 How interesting. 633 00:29:23,505 --> 00:29:27,005 (upbeat orchestral music) 634 00:29:29,690 --> 00:29:32,830 The world's first computer generated musical 635 00:29:32,830 --> 00:29:36,633 will soon take to the stage amid London's West End giants. 636 00:29:38,380 --> 00:29:40,580 No machine currently exists that can 637 00:29:40,580 --> 00:29:42,600 write convincing dialogue. 638 00:29:42,600 --> 00:29:44,993 But that's not the case with music. 639 00:29:47,280 --> 00:29:49,690 Musician Catherine Woolley will now perform 640 00:29:49,690 --> 00:29:53,240 a world premier of our computer's compositions, 641 00:29:53,240 --> 00:29:56,680 created especially for the Experiment. 642 00:29:56,680 --> 00:29:57,850 Okay. 643 00:29:57,850 --> 00:29:58,683 So this is the very first one 644 00:29:58,683 --> 00:30:00,220 that the computer sent through. 645 00:30:01,733 --> 00:30:04,816 (gentle piano music) 646 00:30:10,197 --> 00:30:12,840 There are 200 in the first batch, 647 00:30:12,840 --> 00:30:15,183 each generated in a matter of seconds. 648 00:30:41,970 --> 00:30:45,450 The computer, it would seem, is off to a good start. 649 00:30:45,450 --> 00:30:47,670 This is number 6 of the second batch. 650 00:30:47,670 --> 00:30:49,870 The team reviews several numbers, 651 00:30:49,870 --> 00:30:51,830 looking for tunes to suit everything 652 00:30:51,830 --> 00:30:54,353 from showstoppers to love songs. 653 00:30:58,050 --> 00:30:59,270 Give us the same tune but with 654 00:30:59,270 --> 00:31:00,683 something completely different. 655 00:31:00,683 --> 00:31:03,933 (pleasant piano music) 656 00:31:06,340 --> 00:31:07,173 So, again it's-- 657 00:31:07,173 --> 00:31:09,000 It's a beautiful delicacy. 658 00:31:09,000 --> 00:31:11,270 I didn't know if we were gonna get 659 00:31:11,270 --> 00:31:12,750 some random sort of like-- 660 00:31:12,750 --> 00:31:14,260 Discordant. Yeah. 661 00:31:14,260 --> 00:31:15,640 And just sort of sitting there going, 662 00:31:15,640 --> 00:31:18,690 I can hear elements of a musical there, 663 00:31:18,690 --> 00:31:20,270 which is quite exciting. 664 00:31:20,270 --> 00:31:22,060 As the composer in the room, 665 00:31:22,060 --> 00:31:24,550 Ben pulls no punches. 666 00:31:24,550 --> 00:31:27,360 I thought that was a shockingly awful melody. 667 00:31:27,360 --> 00:31:31,060 This is not music that is good enough to be in a show. 668 00:31:31,060 --> 00:31:34,360 This has to be distilled, properly distilled, 669 00:31:34,360 --> 00:31:37,490 by a human being to actually turn that 670 00:31:37,490 --> 00:31:41,120 into something that will be palatable in a musical. 671 00:31:41,120 --> 00:31:42,010 But what we have to do, 672 00:31:42,010 --> 00:31:46,000 is we have to honor what we signed up for. 673 00:31:46,000 --> 00:31:47,300 What we can't now do is go, 674 00:31:47,300 --> 00:31:48,757 we've had enough of the computer, 675 00:31:48,757 --> 00:31:51,000 and we're just gonna go a separate way. 676 00:31:51,000 --> 00:31:54,870 So we now have to find a way of navigating what we've got. 677 00:31:54,870 --> 00:31:57,500 In order to do that, you have to invest. 678 00:31:57,500 --> 00:31:59,390 You have to invest in it creatively. 679 00:31:59,390 --> 00:32:03,420 I can't lie, I'm reasonably reassured. 680 00:32:03,420 --> 00:32:04,550 I thought I was gonna come in here 681 00:32:04,550 --> 00:32:07,900 and listen to some really discordant, bizarre stuff. 682 00:32:07,900 --> 00:32:08,930 This is not a show. 683 00:32:08,930 --> 00:32:11,300 There aren't songs there yet, but there's something. 684 00:32:11,300 --> 00:32:12,200 There's something. 685 00:32:14,470 --> 00:32:16,130 What's happening with this musical 686 00:32:16,130 --> 00:32:17,940 is really quite a challenge. 687 00:32:17,940 --> 00:32:19,710 Computer programs are being asked 688 00:32:19,710 --> 00:32:22,800 to deliver something at a professional level. 689 00:32:22,800 --> 00:32:23,950 In some ways it's a little bit cruel, 690 00:32:23,950 --> 00:32:26,155 because we've only really been looking at 691 00:32:26,155 --> 00:32:29,280 how computers can be creative for a short amount of time, 692 00:32:29,280 --> 00:32:32,790 whereas people have been being creative for centuries. 693 00:32:32,790 --> 00:32:35,910 As the field of computational creativity progresses, 694 00:32:35,910 --> 00:32:39,237 and it does raise some interesting philosophical questions, 695 00:32:39,237 --> 00:32:41,940 we will see an increasing amount of work 696 00:32:41,940 --> 00:32:44,150 which is exploring the relationship 697 00:32:44,150 --> 00:32:47,470 between the human creative and the machine. 698 00:32:47,470 --> 00:32:50,680 And then the question becomes, how much agency, 699 00:32:50,680 --> 00:32:53,050 how much autonomy, do we want to give 700 00:32:53,050 --> 00:32:55,228 the machine in the creative process? 701 00:32:55,228 --> 00:32:58,145 (mysterious music) 702 00:32:59,270 --> 00:33:02,191 As a writer of music, you absolutely 703 00:33:02,191 --> 00:33:05,050 are going to want it to be awful. 704 00:33:05,050 --> 00:33:06,917 Absolutely, because the, 705 00:33:10,170 --> 00:33:14,020 the end result of it being wonderful, 706 00:33:14,020 --> 00:33:19,020 or too wonderful, is actually unacceptable. 707 00:33:19,040 --> 00:33:22,650 Therefore, maybe I was a little more 708 00:33:22,650 --> 00:33:25,640 harsh on it than I possibly could have been. 709 00:33:25,640 --> 00:33:27,080 And we fell into the trap 710 00:33:27,080 --> 00:33:28,830 of falling in love with the piece, 711 00:33:28,830 --> 00:33:33,213 and naturally, wanting and needing to take ownership of it. 712 00:33:33,213 --> 00:33:35,480 It's not ours to play with in many respects, 713 00:33:35,480 --> 00:33:38,865 we're a big part of it, but obviously 714 00:33:38,865 --> 00:33:42,089 the Experiment is what's going on here. 715 00:33:42,089 --> 00:33:45,440 (uplifting music) 716 00:33:45,440 --> 00:33:48,210 The Experiment demands that Ben and Nathan, 717 00:33:48,210 --> 00:33:49,930 with Catherine's assistance, 718 00:33:49,930 --> 00:33:51,420 begin to work their way through 719 00:33:51,420 --> 00:33:54,691 our computer composer's prolific output. 720 00:33:54,691 --> 00:33:56,630 (Ben vocalizing) 721 00:33:56,630 --> 00:33:57,870 That's quite nice. 722 00:33:57,870 --> 00:34:01,270 A great song has to get you very quickly. 723 00:34:01,270 --> 00:34:03,500 (Catherine laughs) 724 00:34:03,500 --> 00:34:05,570 In a musical theater song, often there's a verse, 725 00:34:05,570 --> 00:34:09,300 a short intro of some kind, that is intriguing. 726 00:34:09,300 --> 00:34:11,220 It needs to tease you, it needs to say, 727 00:34:11,220 --> 00:34:12,720 I'm onto something interesting, 728 00:34:12,720 --> 00:34:14,980 I'm going somewhere, stay with me, stay with me. 729 00:34:14,980 --> 00:34:17,294 This is the phrase I'm interested in. 730 00:34:17,294 --> 00:34:18,960 (Catherine plays piano notes) 731 00:34:18,960 --> 00:34:23,730 A hook is smaller than a chorus, smaller than a verse. 732 00:34:23,730 --> 00:34:26,810 It's just a little thing that get you. 733 00:34:26,810 --> 00:34:28,272 I mean, that's really nice to-- 734 00:34:28,272 --> 00:34:29,105 It is. 735 00:34:31,820 --> 00:34:35,536 Jesus Christ Superstar begins with that fantastic, 736 00:34:35,536 --> 00:34:38,060 ♪ Bom bom bom, bom pa dum pa dum ♪ 737 00:34:38,060 --> 00:34:40,376 ♪ Bom bom bom, bom pa dum pa dum ♪ 738 00:34:40,376 --> 00:34:44,793 Immediately I found myself, when I first heard that, going. 739 00:34:47,460 --> 00:34:50,213 Weeks pass and progress is slow. 740 00:34:51,450 --> 00:34:56,050 Android's songs deliver melodies, but they lack structure. 741 00:34:56,050 --> 00:34:57,510 I want to know that it doesn't just 742 00:34:57,510 --> 00:35:00,137 keep returning to, ba ba ba, ba ba ba ba. 743 00:35:01,190 --> 00:35:02,420 Just something that goes somewhere. 744 00:35:02,420 --> 00:35:04,100 I know. Takes you somewhere. 745 00:35:04,100 --> 00:35:06,130 Android doesn't quite come up 746 00:35:06,130 --> 00:35:10,110 with really neatly, metrically packaged songs. 747 00:35:10,110 --> 00:35:13,070 It tends to be a bit more freeform in what it does. 748 00:35:13,070 --> 00:35:15,620 But there's a really valuable research experiment 749 00:35:15,620 --> 00:35:17,383 that's being conducted here. 750 00:35:17,383 --> 00:35:21,080 (gentle piano music) 751 00:35:21,080 --> 00:35:23,260 From the music our team finds, 752 00:35:23,260 --> 00:35:26,253 they must build 16 full songs. 753 00:35:29,367 --> 00:35:30,310 I'm losing what's left, 754 00:35:30,310 --> 00:35:32,290 honestly. Yeah, me too, actually. 755 00:35:32,290 --> 00:35:36,080 We have only got through 229 bars out of 756 00:35:36,957 --> 00:35:38,584 1,488. But already my brain feels 757 00:35:38,584 --> 00:35:40,300 like I can't filter anything. 758 00:35:40,300 --> 00:35:41,810 We've got two weeks before we go 759 00:35:41,810 --> 00:35:43,460 into rehearsals with a team of actors, 760 00:35:43,460 --> 00:35:45,293 and there is nothing to play with. 761 00:35:47,770 --> 00:35:48,640 And working with 762 00:35:48,640 --> 00:35:51,423 Android's music is just the start. 763 00:35:52,853 --> 00:35:53,860 ♪ I'm calm, I'm cool ♪ 764 00:35:53,860 --> 00:35:54,850 ♪ A gibbering fool ♪ 765 00:35:54,850 --> 00:35:57,030 ♪ Is something I'd never become ♪ 766 00:35:57,030 --> 00:35:58,069 ♪ When thunder is rumbling ♪ 767 00:35:58,069 --> 00:35:59,147 ♪ And others are crumbling ♪ 768 00:35:59,147 --> 00:36:01,110 ♪ I hum, mm-hmm ♪ 769 00:36:01,110 --> 00:36:02,950 So now you're a song with you 770 00:36:02,950 --> 00:36:05,223 the answer's like that and you're in love. 771 00:36:06,760 --> 00:36:09,650 That's not a coherent sentence. 772 00:36:09,650 --> 00:36:11,180 Ben and Nathan are scouring 773 00:36:11,180 --> 00:36:14,060 the handiwork of the Cloud Lyricist, 774 00:36:14,060 --> 00:36:16,560 a computer system built for the Experiment 775 00:36:16,560 --> 00:36:18,650 that writes lyrics. 776 00:36:18,650 --> 00:36:23,193 Its first hard copy submission numbers 20,000 lines. 777 00:36:24,250 --> 00:36:29,250 By reading 7,000 musical theater songs, it learned to write. 778 00:36:29,450 --> 00:36:31,970 But not quite how we would. 779 00:36:31,970 --> 00:36:34,060 Individually, they all make sense. 780 00:36:34,060 --> 00:36:36,810 I want our moment, and I will love you. 781 00:36:36,810 --> 00:36:39,100 But what if you can't see. 782 00:36:39,100 --> 00:36:42,080 I give into facts and life is beautiful. 783 00:36:42,080 --> 00:36:46,740 I have seen the love and I have to know, is my phone. 784 00:36:46,740 --> 00:36:50,210 Even if they give you a stroke of absolute genius, 785 00:36:50,210 --> 00:36:52,120 they can never follow that up 786 00:36:52,120 --> 00:36:54,173 with anything that's associated to that. 787 00:36:55,100 --> 00:36:57,100 Whether that's a melody that develops 788 00:36:57,100 --> 00:36:58,880 like a normal melody would develop, 789 00:36:58,880 --> 00:37:00,520 or whether it's a sentence that develops 790 00:37:00,520 --> 00:37:02,220 like a normal sentence would develop. 791 00:37:02,220 --> 00:37:06,600 They do not think in the same way as human beings. 792 00:37:06,600 --> 00:37:08,743 Every new thought is a new thought. 793 00:37:10,700 --> 00:37:12,560 It's for this reason that Ben and Nathan 794 00:37:12,560 --> 00:37:14,810 have had to write the script for the musical. 795 00:37:15,874 --> 00:37:18,210 But the Experiment requires that they incorporate 796 00:37:18,210 --> 00:37:20,340 as many computer generated lyrics 797 00:37:20,340 --> 00:37:22,443 as possible into the new songs. 798 00:37:24,280 --> 00:37:26,740 In the future, as these systems improve, 799 00:37:26,740 --> 00:37:28,230 they'll be able to write longer 800 00:37:28,230 --> 00:37:29,878 and longer stretches of text. 801 00:37:29,878 --> 00:37:32,520 And eventually, there's no fundamental reason 802 00:37:32,520 --> 00:37:33,700 why they wouldn't be able to generate 803 00:37:33,700 --> 00:37:36,620 entire stories, entire songs. 804 00:37:36,620 --> 00:37:38,930 However, the question would then be, 805 00:37:38,930 --> 00:37:40,530 what do we want to use those systems for? 806 00:37:40,530 --> 00:37:44,360 Would we really want to generate a musical every second, 807 00:37:44,360 --> 00:37:47,680 or perhaps do we use these systems in a different way, 808 00:37:47,680 --> 00:37:49,430 to potentially create new art forms 809 00:37:49,430 --> 00:37:51,682 that we haven't even thought of yet. 810 00:37:51,682 --> 00:37:54,030 ♪ Another opening, another show ♪ 811 00:37:54,030 --> 00:37:56,500 With just three months until curtain up, 812 00:37:56,500 --> 00:37:58,873 the Experiment must now shift up a gear. 813 00:38:00,023 --> 00:38:03,020 ♪ Another opening of another show ♪ 814 00:38:03,020 --> 00:38:04,470 A team of actors and creatives 815 00:38:04,470 --> 00:38:07,840 have gathered to workshop what's been written so far, 816 00:38:07,840 --> 00:38:11,689 and to test how it goes down with a real, live audience. 817 00:38:11,689 --> 00:38:14,010 ♪ Another opening of another show ♪ 818 00:38:14,010 --> 00:38:15,340 We've written six songs now. 819 00:38:15,340 --> 00:38:16,990 We've got a whole script. 820 00:38:16,990 --> 00:38:19,100 We've got a whole cast of actors. 821 00:38:19,100 --> 00:38:20,660 So we're really excited. 822 00:38:20,660 --> 00:38:22,800 You know, by the end of this week, 823 00:38:22,800 --> 00:38:26,333 we will have worked out whether or not this is any good. 824 00:38:26,333 --> 00:38:29,130 ♪ Another opening of another show ♪ 825 00:38:29,130 --> 00:38:30,910 The cast face the week ahead 826 00:38:30,910 --> 00:38:34,600 with no idea where this show has come from. 827 00:38:34,600 --> 00:38:37,780 Because this is unlike any other show in history. 828 00:38:37,780 --> 00:38:40,500 And the reason it's unlike any other show in history 829 00:38:40,500 --> 00:38:44,860 is because it is the first piece 830 00:38:44,860 --> 00:38:48,193 of musical theater generated by computers. 831 00:38:49,811 --> 00:38:50,701 Cool! Yeah. 832 00:38:50,701 --> 00:38:52,000 (laughing) 833 00:38:52,000 --> 00:38:55,310 Everything from the premise, the plot, 834 00:38:55,310 --> 00:38:58,200 the characters, the structure, the setting, 835 00:38:58,200 --> 00:39:01,010 has come from computer technology. 836 00:39:01,010 --> 00:39:04,565 Now, has anyone got any initial responses or questions? 837 00:39:04,565 --> 00:39:05,398 (laughing) 838 00:39:05,398 --> 00:39:07,843 Was there computer generated casting? 839 00:39:07,843 --> 00:39:11,020 No, you are all here as artists, 840 00:39:11,020 --> 00:39:13,410 and you're here because we think 841 00:39:13,410 --> 00:39:15,728 you're the right people to shape the material. 842 00:39:15,728 --> 00:39:18,320 (vocalizing) 843 00:39:18,320 --> 00:39:19,650 The new team are warming up 844 00:39:19,650 --> 00:39:22,513 for a vocal challenge like none other before. 845 00:39:24,410 --> 00:39:25,970 They're about to be given a taste 846 00:39:25,970 --> 00:39:30,473 of a computer generated song, unseen by Ben and Nathan. 847 00:39:32,170 --> 00:39:34,220 We wanted to look at something 848 00:39:34,220 --> 00:39:36,660 in its purest form from a computer. 849 00:39:36,660 --> 00:39:38,550 So we have a piece of material. 850 00:39:38,550 --> 00:39:40,721 It's by Android Lloyd Webber. 851 00:39:40,721 --> 00:39:42,600 (laughing) 852 00:39:42,600 --> 00:39:43,740 For one time only, 853 00:39:43,740 --> 00:39:47,520 the company will learn a song that is pure computer music, 854 00:39:47,520 --> 00:39:49,990 composed to pure computer lyrics. 855 00:39:49,990 --> 00:39:52,840 So press the button, and this is what comes out. 856 00:39:53,780 --> 00:39:55,423 So it goes, one, two. 857 00:39:56,558 --> 00:39:59,917 ♪ Time is on the tree ♪ 858 00:39:59,917 --> 00:40:04,746 ♪ He's a man who can shame ♪ 859 00:40:04,746 --> 00:40:09,746 ♪ A shadow of melody ♪ 860 00:40:11,268 --> 00:40:14,520 ♪ Me has so that the shade ♪ 861 00:40:14,520 --> 00:40:16,730 The Experiment shows how hard it is 862 00:40:16,730 --> 00:40:19,920 for a computer to write lyrics or music, 863 00:40:19,920 --> 00:40:22,433 and combining the two is even harder. 864 00:40:23,350 --> 00:40:26,810 Instead, it's our collaboration between computers 865 00:40:26,810 --> 00:40:29,596 and humans that may get us our hit. 866 00:40:29,596 --> 00:40:31,950 ♪ Now there's only tomorrow ♪ 867 00:40:31,950 --> 00:40:33,580 Two of Android's melodies, 868 00:40:33,580 --> 00:40:36,550 found in Ben and Nathan's creative garret, 869 00:40:36,550 --> 00:40:40,062 form the heart of the first song the company will learn. 870 00:40:40,062 --> 00:40:40,895 One, two. 871 00:40:40,895 --> 00:40:43,984 ♪ At our feet ♪ 872 00:40:43,984 --> 00:40:46,456 ♪ Stop their hating ♪ 873 00:40:46,456 --> 00:40:47,980 ♪ Show them now ♪ 874 00:40:47,980 --> 00:40:49,940 With one in five of its lines 875 00:40:49,940 --> 00:40:53,320 from the Cloud Lyricist, Ben and Nathan have crafted 876 00:40:53,320 --> 00:40:56,135 something a bit closer to what's needed. 877 00:40:56,135 --> 00:41:00,142 ♪ It's time to show them now ♪ 878 00:41:00,142 --> 00:41:02,809 ♪ We are strong ♪ 879 00:41:04,361 --> 00:41:07,380 That's it, yeah. Come on, good. 880 00:41:07,380 --> 00:41:11,078 ♪ We shall not be moved ♪ 881 00:41:11,078 --> 00:41:13,363 ♪ Hear our song ♪ 882 00:41:13,363 --> 00:41:16,020 I think the reason people involved in theater 883 00:41:16,020 --> 00:41:18,470 do what they do is because they are human beings, 884 00:41:18,470 --> 00:41:21,400 and they have a human emotional connection 885 00:41:21,400 --> 00:41:25,270 to the story, to the characters, to the songs, to the dance. 886 00:41:25,270 --> 00:41:28,240 And I just can't imagine how a computer, 887 00:41:28,240 --> 00:41:30,050 I mean, prove me wrong. 888 00:41:30,050 --> 00:41:32,580 I can't imagine how a computer is gonna move me. 889 00:41:32,580 --> 00:41:34,240 The feeling, the pain and the pathos, 890 00:41:34,240 --> 00:41:36,059 or the humor of a character. 891 00:41:36,059 --> 00:41:38,430 ♪ You make it demonstrating ♪ 892 00:41:38,430 --> 00:41:40,760 With its strange computer origins, 893 00:41:40,760 --> 00:41:42,370 Beyond the Fence centers around 894 00:41:42,370 --> 00:41:44,230 the Greenham Common anti-nuclear 895 00:41:44,230 --> 00:41:47,133 Women's Peace Camp in the early 1980s. 896 00:41:48,980 --> 00:41:52,840 It's a computer generated quest in which soldier Jim 897 00:41:52,840 --> 00:41:55,890 must learn to understand a child, George. 898 00:41:55,890 --> 00:41:59,563 Leading to an unlikely alliance with protester Mary. 899 00:42:00,404 --> 00:42:03,526 ♪ And I'm so scared of that sound ♪ 900 00:42:03,526 --> 00:42:06,010 ♪ Scared of that sound ♪ 901 00:42:06,010 --> 00:42:09,570 The music is revealing some difficulties. 902 00:42:09,570 --> 00:42:10,960 Just read the music. 903 00:42:10,960 --> 00:42:11,977 Just stick to the. 904 00:42:13,600 --> 00:42:15,230 What's interesting, there's lots 905 00:42:15,230 --> 00:42:18,100 of familiar elements there, 906 00:42:18,100 --> 00:42:20,240 but there's just weird things 907 00:42:20,240 --> 00:42:22,953 happening with rhythm and timing. 908 00:42:22,953 --> 00:42:25,083 ♪ And I'm so scared of that ♪ 909 00:42:25,083 --> 00:42:28,618 ♪ Scared of that sound ♪ 910 00:42:28,618 --> 00:42:32,140 ♪ Just a bunch of tie-dye terrorists ♪ 911 00:42:32,140 --> 00:42:35,400 The lyrics are presenting their own challenges, 912 00:42:35,400 --> 00:42:38,197 and every line is scrutinized. 913 00:42:38,197 --> 00:42:39,979 ♪ It is my job ♪ 914 00:42:39,979 --> 00:42:41,959 ♪ It is my duty ♪ 915 00:42:41,959 --> 00:42:44,180 Of the 74 lines of this song, 916 00:42:44,180 --> 00:42:46,470 14 were written by the Experiment's 917 00:42:46,470 --> 00:42:49,946 computer lyricist, nicknamed Clarissa. 918 00:42:49,946 --> 00:42:52,775 ♪ It's a brand new kind of salvation ♪ 919 00:42:52,775 --> 00:42:55,510 ♪ For their broken nation ♪ 920 00:42:55,510 --> 00:42:56,550 Great, lovely. 921 00:42:56,550 --> 00:42:57,594 We don't have, it's a brand new 922 00:42:57,594 --> 00:43:00,570 kind of salvation, we have just, 923 00:43:00,570 --> 00:43:02,664 it's a brand new salvation. Oh yeah, sorry. 924 00:43:02,664 --> 00:43:04,550 A brand new salvation comes from the computer. 925 00:43:04,550 --> 00:43:05,520 So did the computer give us, 926 00:43:05,520 --> 00:43:07,540 it's a brand new kind of salvation, or-- 927 00:43:07,540 --> 00:43:08,763 Actually no, it just gives-- 928 00:43:08,763 --> 00:43:10,590 It gave us a brand new salvation, 929 00:43:10,590 --> 00:43:13,550 but we needed extra syllables to make the line work. 930 00:43:13,550 --> 00:43:15,590 Two extra words were added 931 00:43:15,590 --> 00:43:18,890 to fit the computer's lyric to the rhythm of the music. 932 00:43:18,890 --> 00:43:21,050 So if we did the purist computer version, 933 00:43:21,050 --> 00:43:24,350 it's a brand new salvation, without 934 00:43:24,350 --> 00:43:26,760 the extra syllables, how does that scan? 935 00:43:26,760 --> 00:43:27,826 It would be, 936 00:43:27,826 --> 00:43:30,130 ♪ It's a brand new salvation ♪ 937 00:43:30,130 --> 00:43:31,370 I can't allow that. 938 00:43:31,370 --> 00:43:34,610 I'm such a purist about Wester's land, I can't. 939 00:43:34,610 --> 00:43:36,340 It's a brand new salvation. It has to be 940 00:43:36,340 --> 00:43:37,280 a brand new one, yeah. 941 00:43:37,280 --> 00:43:39,032 It's a brand new salvation. 942 00:43:39,032 --> 00:43:40,946 It's a brand new salvation. 943 00:43:40,946 --> 00:43:42,537 ♪ It's a brand new salvation ♪ 944 00:43:42,537 --> 00:43:46,260 ♪ It's a brand new salvation ♪ 945 00:43:46,260 --> 00:43:47,910 What would you feel about that? 946 00:43:49,400 --> 00:43:51,446 I feel less upset about that. 947 00:43:51,446 --> 00:43:54,797 ♪ It's a brand new salvation ♪ 948 00:43:54,797 --> 00:43:56,163 I like that. Yeah? 949 00:43:57,600 --> 00:43:58,785 I think Clarissa would be happy with that. 950 00:43:58,785 --> 00:44:00,656 Yeah, I think she'd be happy with that. 951 00:44:00,656 --> 00:44:03,745 ♪ They know that this deterrent keeps us free ♪ 952 00:44:03,745 --> 00:44:06,856 ♪ It's a brand new salvation ♪ 953 00:44:06,856 --> 00:44:09,374 ♪ For their broken nation ♪ 954 00:44:09,374 --> 00:44:12,499 ♪ You've gotta get those bums on seats ♪ 955 00:44:12,499 --> 00:44:15,367 ♪ That's what it's all about ♪ 956 00:44:15,367 --> 00:44:17,500 The Experiment now demands 957 00:44:17,500 --> 00:44:20,190 a rigorous test of the rehearsed material, 958 00:44:20,190 --> 00:44:22,321 which means finding an audience. 959 00:44:22,321 --> 00:44:25,300 ♪ An empty theater means mountainous debts ♪ 960 00:44:25,300 --> 00:44:26,460 We've been granted access 961 00:44:26,460 --> 00:44:29,310 to the database of Delfont Mackintosh Theatres, 962 00:44:29,310 --> 00:44:32,700 who own eight of the big West End stages. 963 00:44:32,700 --> 00:44:34,760 Using an algorithm, of course, 964 00:44:34,760 --> 00:44:38,173 we've identified their most ardent musical fans. 965 00:44:40,110 --> 00:44:41,850 Our very discerning guinea pigs 966 00:44:41,850 --> 00:44:44,885 have no idea what they're about to see. 967 00:44:44,885 --> 00:44:47,856 ♪ Every bum is a bum you could kiss ♪ 968 00:44:47,856 --> 00:44:49,480 ♪ Get 'em in, sit 'em down ♪ 969 00:44:49,480 --> 00:44:54,480 ♪ You just gotta get those bums on seats ♪ 970 00:44:54,571 --> 00:44:55,520 Good evening, ladies and gentlemen, 971 00:44:55,520 --> 00:44:57,520 and thank you for coming along this evening. 972 00:44:57,520 --> 00:44:59,370 This is the first time that this material 973 00:44:59,370 --> 00:45:03,350 has ever been performed live to a real live audience. 974 00:45:03,350 --> 00:45:05,430 The computer's recipe, idea, 975 00:45:05,430 --> 00:45:08,812 music, and lyrics will be put to the test. 976 00:45:08,812 --> 00:45:11,155 ♪ Every time we stop a vehicle ♪ 977 00:45:11,155 --> 00:45:14,250 ♪ From entering that base ♪ 978 00:45:14,250 --> 00:45:16,330 The showing begins in the peace camp, 979 00:45:16,330 --> 00:45:19,793 with a number to induce high energy, but low tension. 980 00:45:21,330 --> 00:45:24,320 Bonds between characters are strengthened. 981 00:45:24,320 --> 00:45:26,503 Your mother is not in love with me. 982 00:45:27,340 --> 00:45:28,907 Come on. 983 00:45:28,907 --> 00:45:30,883 But I would like us to be friends. 984 00:45:32,010 --> 00:45:34,360 Do you think that's possible? 985 00:45:34,360 --> 00:45:37,000 As tenderness should peak mid-act, 986 00:45:37,000 --> 00:45:39,040 our soldier and female lead come 987 00:45:39,040 --> 00:45:40,659 a step closer to finding true love. 988 00:45:40,659 --> 00:45:43,008 ♪ And I'm so scared of that ♪ 989 00:45:43,008 --> 00:45:48,008 ♪ Scared of that sound ♪ 990 00:45:49,149 --> 00:45:51,760 ♪ Nothing can stop us now ♪ 991 00:45:51,760 --> 00:45:54,290 And finally, we reach At Our Feet, 992 00:45:54,290 --> 00:45:56,773 the high-energy act one closer. 993 00:45:58,499 --> 00:46:03,499 ♪ The future lies on ♪ 994 00:46:05,755 --> 00:46:08,255 ♪ At our feet ♪ 995 00:46:12,506 --> 00:46:15,506 (audience applauds) 996 00:46:21,908 --> 00:46:24,257 Thank you very much. 997 00:46:24,257 --> 00:46:25,640 So, is there anyone who sort of has 998 00:46:25,640 --> 00:46:27,390 anything that they wanted to share? 999 00:46:27,390 --> 00:46:28,690 The harmonies we really loved, 1000 00:46:28,690 --> 00:46:30,240 'cause they worked so well together. 1001 00:46:30,240 --> 00:46:31,450 For me, the characters come alive 1002 00:46:31,450 --> 00:46:33,260 and they grew a lot for me through the songs, 1003 00:46:33,260 --> 00:46:35,190 which is something I like in a musical. 1004 00:46:35,190 --> 00:46:38,521 If you were to imagine the people who wrote this, 1005 00:46:38,521 --> 00:46:41,163 do you have a sort of picture of them in your mind at all? 1006 00:46:42,435 --> 00:46:45,610 I felt I recognized some of the melodies 1007 00:46:45,610 --> 00:46:47,120 and some of the tunes. 1008 00:46:47,120 --> 00:46:48,640 It felt really new and fresh 1009 00:46:48,640 --> 00:46:52,760 and the music didn't sound like anything I'd heard. 1010 00:46:52,760 --> 00:46:57,760 So, I'm afraid we haven't been entirely honest with you. 1011 00:46:57,980 --> 00:47:00,810 And there is one other piece of information 1012 00:47:00,810 --> 00:47:02,790 which I would like to share with you. 1013 00:47:02,790 --> 00:47:05,300 What you have just witnessed is the first, 1014 00:47:05,300 --> 00:47:07,440 to our knowledge, in history, 1015 00:47:07,440 --> 00:47:12,093 the first computer generated musical. 1016 00:47:16,150 --> 00:47:18,110 So if I were sat in your position. 1017 00:47:18,110 --> 00:47:21,043 (audience laughing) 1018 00:47:21,043 --> 00:47:24,220 I think I would have quite a lot of questions. 1019 00:47:24,220 --> 00:47:25,760 I didn't know what to think. 1020 00:47:25,760 --> 00:47:28,110 I was so confused, I didn't know what he meant. 1021 00:47:29,530 --> 00:47:31,160 What even is that? 1022 00:47:31,160 --> 00:47:34,130 But I think it's so intriguing. 1023 00:47:34,130 --> 00:47:38,010 And it's created a whole new way of making a musical, 1024 00:47:38,010 --> 00:47:40,463 that has got infinite possibilities. 1025 00:47:41,430 --> 00:47:42,263 Wow. 1026 00:47:42,263 --> 00:47:43,410 It's just mind-blowing. 1027 00:47:43,410 --> 00:47:44,487 I'm so excited about coming to see it. 1028 00:47:44,487 --> 00:47:47,290 I definitely think it'll be big. 1029 00:47:47,290 --> 00:47:48,123 Definitely. 1030 00:47:49,610 --> 00:47:51,307 So far, so good. 1031 00:47:51,307 --> 00:47:54,110 But the Experiment is not done yet, 1032 00:47:54,110 --> 00:47:56,990 and with nine songs still to be written, 1033 00:47:56,990 --> 00:47:59,643 a new computer is soon to join the party. 1034 00:48:00,796 --> 00:48:02,810 (guitar music) 1035 00:48:02,810 --> 00:48:05,580 Dr. Francois Pachet is a jazz musician 1036 00:48:05,580 --> 00:48:07,363 and computer scientist. 1037 00:48:09,000 --> 00:48:10,930 Having heard about the Experiment, 1038 00:48:10,930 --> 00:48:12,450 he has offered us the chance to use 1039 00:48:12,450 --> 00:48:15,050 a revolutionary new system called 1040 00:48:15,050 --> 00:48:17,910 Flow Composer for the final push. 1041 00:48:17,910 --> 00:48:20,260 Flow Composer, it's an interactive system. 1042 00:48:20,260 --> 00:48:23,220 It's not a generation system where you would 1043 00:48:23,220 --> 00:48:25,053 just click a button and that's it. 1044 00:48:26,520 --> 00:48:28,550 Unlike Android, Flow Composer 1045 00:48:28,550 --> 00:48:31,563 is a specialist in creative conversation. 1046 00:48:32,950 --> 00:48:35,730 To demonstrate the potential of a partnership, 1047 00:48:35,730 --> 00:48:39,927 Francois goes to work on a song being written for the show. 1048 00:48:39,927 --> 00:48:44,010 Okay, so I suggest quite a brutal edit 1049 00:48:44,010 --> 00:48:46,850 and remove the first nine bars. 1050 00:48:46,850 --> 00:48:47,683 Go on. 1051 00:48:47,683 --> 00:48:49,320 Then the computer will suggest 1052 00:48:49,320 --> 00:48:51,903 a brand new stretch of music to fill the gap. 1053 00:48:52,737 --> 00:48:55,487 (pleasant music) 1054 00:49:00,670 --> 00:49:02,490 Suppose you don't like it. 1055 00:49:04,948 --> 00:49:06,445 This is the new version. 1056 00:49:06,445 --> 00:49:09,195 (pleasant music) 1057 00:49:11,720 --> 00:49:14,650 Here you have a wrong note, but you see here, we can-- 1058 00:49:14,650 --> 00:49:17,240 Just make it as natural. 1059 00:49:17,240 --> 00:49:18,610 Okay, here is natural. 1060 00:49:18,610 --> 00:49:20,197 You want better. 1061 00:49:20,197 --> 00:49:22,947 (pleasant music) 1062 00:49:25,952 --> 00:49:28,010 I almost feel guilty when I'm doing 1063 00:49:28,010 --> 00:49:29,120 all of that stuff, though. Oh, yes? 1064 00:49:29,120 --> 00:49:31,073 This is interesting. 1065 00:49:31,073 --> 00:49:34,230 I feel good when I destroy things. 1066 00:49:34,230 --> 00:49:36,370 I worry that that's sort of breaking the rules 1067 00:49:36,370 --> 00:49:39,050 of the game when you go, oh I don't like that, 1068 00:49:39,050 --> 00:49:40,252 I'll just change that line. 1069 00:49:40,252 --> 00:49:42,251 I think it's precisely the rule of the game is to break. 1070 00:49:42,251 --> 00:49:44,810 It's a break and repair game. 1071 00:49:44,810 --> 00:49:48,000 You break, the user, system repairs. 1072 00:49:48,000 --> 00:49:49,820 And because the system is never tired, 1073 00:49:49,820 --> 00:49:51,950 you can do this infinitely. 1074 00:49:53,807 --> 00:49:55,770 Away from the rehearsal room, 1075 00:49:55,770 --> 00:49:59,663 Nathan tackles one of the few remaining songs to be written. 1076 00:50:01,000 --> 00:50:04,130 To collaborate with the new Flow Composer system, 1077 00:50:04,130 --> 00:50:05,820 you first must choose the style 1078 00:50:05,820 --> 00:50:07,270 in which you'd like to write. 1079 00:50:08,370 --> 00:50:11,380 Be it disco, pop, or in this case, 1080 00:50:11,380 --> 00:50:13,463 anti-nuclear protest song. 1081 00:50:14,446 --> 00:50:17,392 ♪ Hand in hand the line extends ♪ 1082 00:50:17,392 --> 00:50:20,237 ♪ All around the nine mile fence ♪ 1083 00:50:20,237 --> 00:50:23,301 ♪ 30,000 women chant ♪ 1084 00:50:23,301 --> 00:50:26,160 ♪ Bring the message home ♪ 1085 00:50:26,160 --> 00:50:27,600 Real songs sung at 1086 00:50:27,600 --> 00:50:30,730 the Greenham Common Peace Camp are analyzed, 1087 00:50:30,730 --> 00:50:32,950 and without sampling or copying, 1088 00:50:32,950 --> 00:50:35,793 the computer creates an entirely new song. 1089 00:50:38,200 --> 00:50:40,533 We Are Greenham is born. 1090 00:50:44,250 --> 00:50:47,080 Flow Composer generated a melody 1091 00:50:47,080 --> 00:50:51,530 that came entirely from the information given to it 1092 00:50:51,530 --> 00:50:55,080 by the women who sang those songs 30 years ago. 1093 00:50:55,080 --> 00:50:57,842 Genuinely think you can hear that in We Are Greenham. 1094 00:50:57,842 --> 00:51:02,509 (piano music) (humming) 1095 00:51:12,680 --> 00:51:14,530 Pause, pause, pause, pause. 1096 00:51:20,111 --> 00:51:24,228 ♪ They can't take away our sorrow ♪ 1097 00:51:24,228 --> 00:51:28,134 ♪ They can't take away our song ♪ 1098 00:51:28,134 --> 00:51:30,119 ♪ If we let them win ♪ 1099 00:51:30,119 --> 00:51:35,119 ♪ We'll all stand in ♪ 1100 00:51:35,276 --> 00:51:39,050 ♪ Which side are you on ♪ 1101 00:51:39,050 --> 00:51:41,240 Greenham Common in Berkshire, 1102 00:51:41,240 --> 00:51:45,053 a former US Air Force base, housing nuclear cruise missiles. 1103 00:51:50,710 --> 00:51:53,940 For 19 years, women came here in the thousands 1104 00:51:53,940 --> 00:51:56,173 to protest against nuclear weapons. 1105 00:51:57,030 --> 00:52:01,003 Many lived here, camping beyond the fence from the silos. 1106 00:52:02,670 --> 00:52:05,803 The songs that everybody knew were the protest songs. 1107 00:52:08,220 --> 00:52:09,600 We wouldn't suddenly sit around 1108 00:52:09,600 --> 00:52:12,380 singing the Beatles, you know. (laughs) 1109 00:52:12,380 --> 00:52:15,130 The songs are all about strength in numbers 1110 00:52:15,130 --> 00:52:18,510 and people power and love, 1111 00:52:18,510 --> 00:52:20,600 and so they're nice things to sing, 1112 00:52:20,600 --> 00:52:24,886 and I think helped to buoy people's spirits at times. 1113 00:52:24,886 --> 00:52:27,320 ♪ We are Greenham ♪ 1114 00:52:27,320 --> 00:52:29,304 ♪ We are home ♪ 1115 00:52:29,304 --> 00:52:31,874 ♪ We are freedom ♪ 1116 00:52:31,874 --> 00:52:34,062 ♪ We are strong ♪ 1117 00:52:34,062 --> 00:52:39,062 ♪ We don't want their Armageddon ♪ 1118 00:52:40,131 --> 00:52:43,420 ♪ They will hear our song ♪ 1119 00:52:43,420 --> 00:52:44,676 I mean, look where we are. 1120 00:52:44,676 --> 00:52:48,300 You couldn't get further away from tapping keyboards. 1121 00:52:48,300 --> 00:52:50,510 Out of everything the computer could've chosen for us, 1122 00:52:50,510 --> 00:52:53,540 we've ended up in the middle of a field, 1123 00:52:53,540 --> 00:52:57,933 surrounded by trees, it's really cold, and that's exciting. 1124 00:52:58,890 --> 00:53:00,040 But it's very far away from 1125 00:53:00,040 --> 00:53:01,783 what I thought we'd ever be given. 1126 00:53:02,820 --> 00:53:03,692 ♪ We are strong ♪ 1127 00:53:03,692 --> 00:53:08,692 ♪ We don't want their Armageddon ♪ 1128 00:53:09,058 --> 00:53:12,558 ♪ They will hear our song ♪ 1129 00:53:14,457 --> 00:53:17,970 (upbeat rock music) 1130 00:53:17,970 --> 00:53:21,220 It's just a week to go until opening night, 1131 00:53:21,220 --> 00:53:22,920 and Beyond the Fence moves into 1132 00:53:22,920 --> 00:53:25,453 the Arts Theatre in London's West End. 1133 00:53:32,750 --> 00:53:36,320 And after a year's research by leading scientists, 1134 00:53:36,320 --> 00:53:40,490 many months of collaboration between humans and computers, 1135 00:53:40,490 --> 00:53:43,103 and six weeks of intense rehearsals, 1136 00:53:45,060 --> 00:53:48,673 the stage is set for the Experiment's final test. 1137 00:53:55,090 --> 00:53:57,060 Of course there's a huge amount unknown with this, 1138 00:53:57,060 --> 00:53:59,360 and although it's a piece that we've all 1139 00:53:59,360 --> 00:54:02,040 been investing in over the last few weeks, 1140 00:54:02,040 --> 00:54:03,840 ultimately this is an experiment, 1141 00:54:03,840 --> 00:54:06,880 and we're doing this for that very specific purpose. 1142 00:54:06,880 --> 00:54:09,630 And who knows what the audience are gonna make of this. 1143 00:54:15,990 --> 00:54:17,480 The experiment to create 1144 00:54:17,480 --> 00:54:21,950 the world's first computer generated musical is complete. 1145 00:54:21,950 --> 00:54:25,170 It's challenged scientists and theater professionals 1146 00:54:25,170 --> 00:54:27,423 to question human creativity. 1147 00:54:28,550 --> 00:54:30,560 To face the blurring boundaries 1148 00:54:30,560 --> 00:54:32,883 between themselves and machines. 1149 00:54:33,790 --> 00:54:36,603 But now, it's the turn of the audience. 1150 00:54:40,042 --> 00:54:42,268 ♪ Opening night, opening night ♪ 1151 00:54:42,268 --> 00:54:45,041 ♪ It's opening night ♪ 1152 00:54:45,041 --> 00:54:45,874 It's gonna happen, and it's gonna happen 1153 00:54:45,874 --> 00:54:47,530 in a very short space of time, 1154 00:54:47,530 --> 00:54:50,680 and there's nothing anyone can do about it now. 1155 00:54:52,385 --> 00:54:54,704 ♪ Will it flop or will it go ♪ 1156 00:54:54,704 --> 00:54:56,430 It doesn't feel like your typical academic project, 1157 00:54:56,430 --> 00:54:58,300 which kind of fizzles out with 1158 00:54:58,300 --> 00:55:00,560 maybe three people reading a paper down the line. 1159 00:55:00,560 --> 00:55:03,078 Instead, you've got real bums on seats. 1160 00:55:03,078 --> 00:55:04,290 ♪ They're on their way ♪ 1161 00:55:04,290 --> 00:55:08,470 A computer is in the same position as anyone 1162 00:55:08,470 --> 00:55:11,510 who's producing or creating a musical. 1163 00:55:11,510 --> 00:55:14,280 But I think the difference is, they're not scared. 1164 00:55:15,727 --> 00:55:17,630 Ladies and gentlemen of the Beyond the Fence company, 1165 00:55:17,630 --> 00:55:19,303 this is your five minute call, your five minute call. 1166 00:55:20,530 --> 00:55:21,860 I'm hoping I'm gonna hate it. 1167 00:55:21,860 --> 00:55:24,870 It's a bit like a chess game against a human and a computer. 1168 00:55:24,870 --> 00:55:27,667 I really, really want the human to win. 1169 00:55:28,903 --> 00:55:30,290 This is the biggest thing that's happened 1170 00:55:30,290 --> 00:55:33,580 to contest creativity in terms of our reach, ever. 1171 00:55:33,580 --> 00:55:37,490 Seeing the two worlds kind of colliding, 1172 00:55:37,490 --> 00:55:40,033 as they have on this, is an extraordinary thing. 1173 00:55:43,420 --> 00:55:46,270 I hope I still have a job tomorrow after seeing it. 1174 00:55:46,270 --> 00:55:49,670 I don't think you can invent the impulse 1175 00:55:49,670 --> 00:55:52,347 that makes you create the work of art. 1176 00:55:52,347 --> 00:55:54,250 Unless a computer one day has 1177 00:55:54,250 --> 00:55:57,693 a heart and a soul, I doubt it can get there. 1178 00:56:00,260 --> 00:56:02,450 Is it gonna get my pulse pounding? 1179 00:56:02,450 --> 00:56:04,730 Is it gonna get me crying? 1180 00:56:04,730 --> 00:56:06,692 It is gonna raise a titter? 1181 00:56:06,692 --> 00:56:08,692 Ooh, just any old day. 1182 00:56:10,412 --> 00:56:11,412 Any old day. 1183 00:56:12,482 --> 00:56:15,360 ♪ What a bum ♪ 1184 00:56:15,360 --> 00:56:17,410 Seven different computer systems 1185 00:56:17,410 --> 00:56:20,163 have shaped every element of Beyond the Fence. 1186 00:56:21,600 --> 00:56:24,693 Our human talent must now give them a voice. 1187 00:56:26,759 --> 00:56:29,926 (dramatic rock music) 1188 00:56:39,920 --> 00:56:44,140 The opening fanfare is unadulterated Android. 1189 00:56:44,140 --> 00:56:47,020 Nick Collins and Francois Pachet's algorithms 1190 00:56:47,020 --> 00:56:49,063 have helped define every song. 1191 00:56:50,084 --> 00:56:53,520 ♪ And again and again and again ♪ 1192 00:56:53,520 --> 00:56:55,960 A huge opening number. 1193 00:56:55,960 --> 00:56:57,800 ♪ Don't waste the war ♪ 1194 00:56:57,800 --> 00:56:59,173 A conflict setting. 1195 00:56:59,173 --> 00:57:02,390 ♪ Don't suffocate the world ♪ 1196 00:57:02,390 --> 00:57:03,680 A female lead. 1197 00:57:03,680 --> 00:57:04,513 A soldier. 1198 00:57:04,513 --> 00:57:07,627 All are choices determined by the data. 1199 00:57:07,627 --> 00:57:10,272 But will they draw the audience in? 1200 00:57:10,272 --> 00:57:14,128 ♪ Our green gate's happening ♪ 1201 00:57:14,128 --> 00:57:16,461 ♪ Happening ♪ 1202 00:57:20,860 --> 00:57:24,630 (audience applauds) 1203 00:57:24,630 --> 00:57:27,350 As the computer generated story unfolds, 1204 00:57:27,350 --> 00:57:30,470 the wounded soldier begins to understand the child, 1205 00:57:30,470 --> 00:57:34,000 whom we learn has not spoken for seven years. 1206 00:57:34,000 --> 00:57:35,971 Hey there, George. 1207 00:57:35,971 --> 00:57:36,804 Happy Halloween. 1208 00:57:38,470 --> 00:57:40,670 I haven't seen you in a while, how you been? 1209 00:57:42,860 --> 00:57:44,860 Almost a quarter of the show's lyrics 1210 00:57:44,860 --> 00:57:48,090 have come from Clarissa, the computer lyricist. 1211 00:57:48,090 --> 00:57:50,460 Four poignant lines from her let Jim 1212 00:57:50,460 --> 00:57:52,704 pour out his heart to George. 1213 00:57:52,704 --> 00:57:56,935 ♪ So swallow me in the shadows ♪ 1214 00:57:56,935 --> 00:58:00,835 ♪ My heart was locked away ♪ 1215 00:58:00,835 --> 00:58:04,033 ♪ I sang one day of the stars ♪ 1216 00:58:04,033 --> 00:58:09,033 ♪ And the stars were blown away ♪ 1217 00:58:12,995 --> 00:58:14,594 ♪ I am graceful ♪ 1218 00:58:14,594 --> 00:58:15,770 ♪ So wonderfully graceful ♪ 1219 00:58:15,770 --> 00:58:16,830 The comedy required 1220 00:58:16,830 --> 00:58:20,034 before the interval arrives on roller skates. 1221 00:58:20,034 --> 00:58:23,586 ♪ I am graceful ♪ 1222 00:58:23,586 --> 00:58:24,868 ♪ So graceful ♪ 1223 00:58:24,868 --> 00:58:26,726 ♪ So bloody else is grateful ♪ 1224 00:58:26,726 --> 00:58:29,226 ♪ To me skids ♪ 1225 00:58:30,942 --> 00:58:33,327 (audience applauds) 1226 00:58:33,327 --> 00:58:35,690 But the real test of the Experiment 1227 00:58:35,690 --> 00:58:37,260 will be whether it actually delivers 1228 00:58:37,260 --> 00:58:41,680 intensity and emotion, heart and soul. 1229 00:58:41,680 --> 00:58:42,513 Bye, George. 1230 00:58:46,770 --> 00:58:47,603 Thank you. 1231 00:58:55,140 --> 00:58:57,010 George, what did you just say? 1232 00:58:57,010 --> 00:58:57,843 Thank you. 1233 00:59:00,041 --> 00:59:01,827 ♪ Thank you ♪ 1234 00:59:01,827 --> 00:59:03,435 ♪ Thank you ♪ 1235 00:59:03,435 --> 00:59:06,204 ♪ For all that you've done ♪ 1236 00:59:06,204 --> 00:59:09,611 ♪ For the gloves and the friendship ♪ 1237 00:59:09,611 --> 00:59:13,331 ♪ The laughter and the gum ♪ 1238 00:59:13,331 --> 00:59:14,730 ♪ Thank you ♪ 1239 00:59:14,730 --> 00:59:16,377 ♪ Thank you ♪ 1240 00:59:16,377 --> 00:59:19,251 ♪ To everyone here ♪ 1241 00:59:19,251 --> 00:59:24,251 ♪ I love you with all of my heart ♪ 1242 00:59:25,444 --> 00:59:30,444 ♪ And the world is a perfect adventure ♪ 1243 00:59:33,701 --> 00:59:37,784 ♪ And I just can't wait to start ♪ 1244 00:59:42,707 --> 00:59:45,707 (audience applauds) 1245 00:59:46,620 --> 00:59:48,700 And after that, it's straight on 1246 00:59:48,700 --> 00:59:51,797 to the finale and the imminent verdict. 1247 00:59:51,797 --> 00:59:56,096 ♪ At our feet ♪ 1248 00:59:56,096 --> 00:59:57,812 ♪ At our feet ♪ 1249 00:59:57,812 --> 00:59:59,833 ♪ Sitting at our feet ♪ 1250 00:59:59,833 --> 01:00:01,982 ♪ At our feet ♪ 1251 01:00:01,982 --> 01:00:04,676 ♪ At our feet ♪ 1252 01:00:04,676 --> 01:00:08,211 ♪ Knowing our future lies ♪ 1253 01:00:08,211 --> 01:00:13,211 ♪ At our feet ♪ 1254 01:00:17,680 --> 01:00:20,180 ♪ At our feet ♪ 1255 01:00:27,897 --> 01:00:30,897 (audience applauds) 1256 01:00:40,515 --> 01:00:42,590 I was really surprised. 1257 01:00:42,590 --> 01:00:43,883 Lots of it works. 1258 01:00:46,090 --> 01:00:47,800 I had to hold in my laughter 1259 01:00:47,800 --> 01:00:49,560 when I recognized Clarissa speaking 1260 01:00:49,560 --> 01:00:52,250 in some times in some of the sadder sections. 1261 01:00:52,250 --> 01:00:53,630 You could definitely note some of the lines 1262 01:00:53,630 --> 01:00:56,030 which were just a tiny bit less human than most. 1263 01:00:57,530 --> 01:00:59,870 I cried, like at the really emotional parts. 1264 01:00:59,870 --> 01:01:00,780 It's just amazing. 1265 01:01:00,780 --> 01:01:03,510 There was a fantastic anthem, wasn't there? 1266 01:01:03,510 --> 01:01:04,343 At Our Feet? 1267 01:01:04,343 --> 01:01:06,143 It was absolutely wonderful. I love it. 1268 01:01:09,010 --> 01:01:10,730 Well I don't think it's gonna entirely 1269 01:01:10,730 --> 01:01:13,663 change the medium or shake things up wildly. 1270 01:01:16,832 --> 01:01:18,010 (laughing) 1271 01:01:18,010 --> 01:01:19,840 I think Elton John, Andrew Lloyd Webber, 1272 01:01:19,840 --> 01:01:22,490 and Stephen Sondheim can sleep easy tonight. 1273 01:01:22,490 --> 01:01:25,630 I thought it was a really great company of actors, 1274 01:01:25,630 --> 01:01:27,400 could really sing, and there was 1275 01:01:27,400 --> 01:01:29,670 the most wonderful little girl. 1276 01:01:29,670 --> 01:01:31,793 But it's got a way to go yet. 1277 01:01:33,700 --> 01:01:34,940 On the red carpet, 1278 01:01:34,940 --> 01:01:37,930 there had been understandable skepticism. 1279 01:01:37,930 --> 01:01:39,220 And now? 1280 01:01:39,220 --> 01:01:44,130 I'm completely in disagreement what I said before. 1281 01:01:44,130 --> 01:01:47,773 That song Thank You by this amazing little girl, 1282 01:01:48,820 --> 01:01:51,740 have as much heart and soul as any song 1283 01:01:52,950 --> 01:01:54,520 any one of us could have written. 1284 01:01:54,520 --> 01:01:56,480 I'm really, really glad we came, 1285 01:01:56,480 --> 01:01:59,431 and I'm really glad to have been part of the experience. 1286 01:01:59,431 --> 01:02:01,130 'Cause actually, it does make you think 1287 01:02:01,130 --> 01:02:04,717 about what you do as a writer, and that's important. 1288 01:02:06,620 --> 01:02:08,170 And what of the scientists 1289 01:02:08,170 --> 01:02:10,670 who've been at the heart of the Experiment? 1290 01:02:10,670 --> 01:02:13,367 To bring together a super team of individuals 1291 01:02:13,367 --> 01:02:16,990 and do something that's insane and crazy at the same time 1292 01:02:16,990 --> 01:02:19,490 is a wonderful result in and of itself. 1293 01:02:19,490 --> 01:02:23,610 We've managed to do something that can contribute, 1294 01:02:23,610 --> 01:02:25,990 even if it's just a very small bit. 1295 01:02:25,990 --> 01:02:27,130 And that's a nice feeling. 1296 01:02:27,130 --> 01:02:29,230 I think this is an extraordinary achievement. 1297 01:02:29,230 --> 01:02:31,970 This mixture of the maths and the computing 1298 01:02:31,970 --> 01:02:34,400 and the creativity and the art 1299 01:02:34,400 --> 01:02:37,150 and the huge human element, because it comes into this. 1300 01:02:37,150 --> 01:02:40,627 So, this is enormously innovative, very exciting, 1301 01:02:40,627 --> 01:02:42,973 and I think it's worked. 1302 01:02:42,973 --> 01:02:46,056 (gentle piano music) 1303 01:02:47,322 --> 01:02:48,970 I don't know whether I was ever skeptical. 1304 01:02:48,970 --> 01:02:50,220 I think I was frightened. 1305 01:02:51,230 --> 01:02:53,790 I had big issues with scientists, 1306 01:02:53,790 --> 01:02:55,160 and I'm frightened of science, 1307 01:02:55,160 --> 01:02:56,870 and I'm frightened of technology, 1308 01:02:56,870 --> 01:02:59,960 and I'm frightened of what that means for me. 1309 01:02:59,960 --> 01:03:01,720 I mean, the interesting thing about this journey 1310 01:03:01,720 --> 01:03:06,720 is that it, on all levels, it's made me address my fears. 1311 01:03:07,100 --> 01:03:09,237 When we started getting bogged down 1312 01:03:09,237 --> 01:03:11,970 with the challenges, I really thought 1313 01:03:11,970 --> 01:03:13,420 there was a chance we'd fail. 1314 01:03:14,310 --> 01:03:15,610 And I don't think we have. 1315 01:03:19,472 --> 01:03:22,719 ♪ Experiment ♪ 1316 01:03:22,719 --> 01:03:27,378 ♪ Make it your motto day and night ♪ 1317 01:03:27,378 --> 01:03:31,107 ♪ Experiment ♪ 1318 01:03:31,107 --> 01:03:35,355 ♪ And it will lead you to the light ♪ 1319 01:03:35,355 --> 01:03:39,404 ♪ The apple on the top of the tree ♪ 1320 01:03:39,404 --> 01:03:43,281 ♪ Is never too high to achieve ♪ 1321 01:03:43,281 --> 01:03:47,072 ♪ So take an example from Eve ♪ 1322 01:03:47,072 --> 01:03:51,358 ♪ Experiment ♪ 1323 01:03:51,358 --> 01:03:54,943 ♪ Be curious ♪ 1324 01:03:54,943 --> 01:03:58,852 ♪ Thought interfering friends may frown ♪ 1325 01:03:58,852 --> 01:04:02,637 ♪ Get furious ♪ 1326 01:04:02,637 --> 01:04:07,371 ♪ At each attempt to hold you down ♪ 1327 01:04:07,371 --> 01:04:11,252 ♪ If this advice you only imply ♪ 1328 01:04:11,252 --> 01:04:16,252 ♪ The future can offer you infinite joy and merriment ♪ 1329 01:04:18,139 --> 01:04:32,085 ♪ Experiment ♪