WEBVTT - The Ghost in the Codec 0:00:15.170 --> 0:00:27.730 Pushkin. A few years ago, I was thinking a lot 0:00:27.770 --> 0:00:33.050 about ghosts and music. I was writing a novel called 0:00:33.090 --> 0:00:36.690 White Tears, about two young music producers who fake in 0:00:36.810 --> 0:00:40.130 nineteen twenties blues record and put it out on the Internet, 0:00:40.570 --> 0:00:44.930 pretending it's the work of a long forgotten musician. Gradually 0:00:44.970 --> 0:00:48.610 they realized that something horrible that's coming towards them out 0:00:48.650 --> 0:00:52.730 of the past. While I was writing, I'd spend hours 0:00:52.770 --> 0:00:56.010 every day listening to recordings made in the years after 0:00:56.050 --> 0:01:02.890 the First World War. The characters in my novel strained 0:01:02.890 --> 0:01:06.770 to hear the sounds of long dead singers and guitarists 0:01:07.450 --> 0:01:11.690 emerging out of the crack and hiss. When you put 0:01:11.730 --> 0:01:15.090 the needle down on a record or press play on 0:01:15.130 --> 0:01:20.490 a sound file, you're inviting a ghost into your room. 0:01:20.650 --> 0:01:25.330 You're inviting a ghost into your head, into the acoustic 0:01:25.410 --> 0:01:30.530 space between your ears. I'm not speaking to you now 0:01:31.570 --> 0:01:35.650 right here on this podcast. I'm not here at all. 0:01:36.770 --> 0:01:44.370 I'm speaking to you from the past. And what if 0:01:44.370 --> 0:01:47.450 mine now, as I talk into a microphone in my 0:01:47.570 --> 0:01:51.690 rigged up COVID recording studio, isn't just splintered off from 0:01:51.690 --> 0:01:55.610 yours by a few weeks or months what if mine 0:01:55.690 --> 0:02:02.010 now is fifty years into your past, or ninety or 0:02:02.050 --> 0:02:08.650 a hundred. Am I dead? Or am I alive? Maybe 0:02:08.650 --> 0:02:17.250 I'm dead, but am I alive? For you? I would 0:02:17.250 --> 0:02:20.170 sit and think about recording and time as I wrote 0:02:20.170 --> 0:02:23.450 my novel with a pair of big clothes back headphones 0:02:23.490 --> 0:02:28.370 clamped over my ears. Late one night, I ran across this. 0:02:37.890 --> 0:02:40.970 Maybe as you hear it now in your now, it 0:02:41.050 --> 0:02:44.890 sounds familiar. Maybe it sounds like something you ought to recognize. 0:02:45.650 --> 0:02:47.850 It certainly tugged at my memory when I heard it, 0:02:48.410 --> 0:02:53.570 like pulling on a thread. The best way I had 0:02:53.610 --> 0:02:56.170 to think about this music was that it sounded like 0:02:56.330 --> 0:03:02.610 the outside of something, something that wasn't there, like a 0:03:02.770 --> 0:03:05.330 cast on a phantom limb, or an empty mold for 0:03:05.370 --> 0:03:10.890 a statue, a vacancy. I was hearing the absence of something. 0:03:22.130 --> 0:03:24.810 This music reminded me of a record that had scared 0:03:24.810 --> 0:03:26.210 the hell out of me when I was a kid. 0:03:26.850 --> 0:03:28.650 It was what they used to call a flexi disc, 0:03:29.130 --> 0:03:31.370 a sheet of thin plastic that came mounted to the 0:03:31.410 --> 0:03:35.970 cover of a magazine called The Unexplained. On it were 0:03:36.050 --> 0:03:40.770 what were portentously called electronic voice phenomena. Supposedly the sounds 0:03:40.770 --> 0:03:43.170 of dead people picked up by various kinds of modern 0:03:43.250 --> 0:03:48.290 audio equipment. The experimental states that, in his opinion, man 0:03:48.450 --> 0:03:52.250 cannot grasp the events after death with his intellect or 0:03:52.290 --> 0:03:57.730 even with his intuition. A voice replies in German air 0:03:57.970 --> 0:04:11.130 can This is into the Zone, a podcast devoted to 0:04:11.170 --> 0:04:15.010 the opposites that shape our world and how borders are 0:04:15.090 --> 0:04:19.250 never as clearly defined as we think. I'm hurry, Kunzru. 0:04:21.090 --> 0:04:24.450 This episode is about something that's fascinated me since I 0:04:24.530 --> 0:04:28.450 heard those voices on the Flexi disc, the ghosts inside 0:04:28.490 --> 0:04:33.930 technology and how they haunt us through signal and noise. 0:04:50.170 --> 0:04:53.810 He can, and we're going to get to signal and 0:04:53.930 --> 0:04:55.890 noise via a place that at first might not seem 0:04:56.010 --> 0:05:00.650 very promising. A conference in Hanover in nineteen eighty eight, 0:05:02.290 --> 0:05:08.050 Hanover Landshah Yes, Hanover, the thirteenth largest city in Germany. 0:05:08.290 --> 0:05:12.410 Hanovers a nineteen eighty eight meeting of the Moving Picture 0:05:12.530 --> 0:05:16.690 Expert screw Up, an organization founded to establish industry wide 0:05:16.730 --> 0:05:25.090 standards for audio and video compression. This is Henover to 0:05:25.170 --> 0:05:29.050 set the scene here in nineteen eighty eight, The coolest 0:05:29.090 --> 0:05:33.050 office technology is the facts. Steve Jobs has been fired 0:05:33.090 --> 0:05:35.450 from Apple for failing to beat IBM in the race 0:05:35.490 --> 0:05:39.130 to develop a personal computer, and Microsoft has released an 0:05:39.210 --> 0:05:43.090 updated version of something called Windows, which is widely considered 0:05:43.090 --> 0:05:46.970 a dud. Your uncle has a new hobby at family parties, 0:05:47.130 --> 0:05:51.050 videotaping everything on his camcorder. The world of music is 0:05:51.090 --> 0:05:56.330 dominated by one futuristic format, the compact disc. Some four 0:05:56.450 --> 0:05:59.610 hundred million CDs will be pressed in nineteen eighty eight. 0:06:00.170 --> 0:06:02.850 You can even get a CD walkman, though a cassette 0:06:02.890 --> 0:06:05.970 tape walkman is still much better for jogging. CD's skip 0:06:06.010 --> 0:06:09.650 if you move them around too much. The audio engineers 0:06:09.650 --> 0:06:12.210 at the hand of a meeting are there to hammer 0:06:12.210 --> 0:06:15.930 out a standard for digital audio, and they're haunted by 0:06:15.970 --> 0:06:20.530 one recent disaster, the video format Wars, in which two 0:06:20.650 --> 0:06:22.930 versions of the same thing went head to head in 0:06:22.930 --> 0:06:29.810 the marketplace, Betamax versus VHS, both chunky plastic shells around 0:06:29.930 --> 0:06:33.730 spools of magnetic tape. With the Betamax at your command, 0:06:34.170 --> 0:06:37.370 you'll never again be deprived of watching whatever program you 0:06:37.450 --> 0:06:41.410 desire at your convenience. You'll be free of the restrictions 0:06:41.450 --> 0:06:44.970 of time. Its uses are defined only by the limits 0:06:45.010 --> 0:06:48.730 of your imagination. But while you might be free of 0:06:48.770 --> 0:06:52.610 the restrictions of time. The beta max wasn't it could 0:06:52.650 --> 0:06:56.130 only record for one hour, but VHS tapes could record 0:06:56.250 --> 0:06:59.330 two hours, long enough for movies that could be rented 0:06:59.370 --> 0:07:04.290 from video stores, providing employment for slackers, stoners and pornographers 0:07:04.370 --> 0:07:09.730 around the world. So VHS one. But the format wars 0:07:09.770 --> 0:07:13.570 caused both sides to hemorrhage cash. When it came to audio, 0:07:14.090 --> 0:07:16.690 no one wanted to risk a repeat of that way 0:07:16.730 --> 0:07:24.130 too stressful. Far better to work out a standard. Yes, 0:07:24.210 --> 0:07:28.570 whole technical should be. I will stop you if I 0:07:28.610 --> 0:07:33.970 think you're getting too technical. How about that? Standardization almost 0:07:33.970 --> 0:07:38.290 by definition, is not something that attracts mavericks or loaners. 0:07:39.050 --> 0:07:42.210 If you're a cowboy or a romantic, go elsewhere with 0:07:42.290 --> 0:07:46.250 your horses and your melodrama. Standardization is a calling for 0:07:46.290 --> 0:07:49.050 people who want to get along, people who like to 0:07:49.090 --> 0:07:51.850 agree on how to do something. People who like to 0:07:51.890 --> 0:07:53.970 make sure that if a thing is going to be done, 0:07:54.370 --> 0:07:57.930 it's done in the best possible way, exactly the same 0:07:58.210 --> 0:08:03.890 by everyone, with no argument, People like Karl Heinz Brandenburg. 0:08:04.530 --> 0:08:11.970 Psych Acoustics is the science merely off aring and viewing 0:08:13.450 --> 0:08:18.930 or ears and brain subleck box in Hanover in nineteen 0:08:18.930 --> 0:08:22.290 eighty eight, Karl Heinz led one of the fourteen teams 0:08:22.290 --> 0:08:27.330 who had competing proposals for an audio compression standard. All 0:08:27.330 --> 0:08:30.890 sorts of companies were interested in audio compression. TV and 0:08:31.010 --> 0:08:38.010 radio stations, telecoms, manufacturers of consumer electronics, governments. Karl Hines 0:08:38.050 --> 0:08:40.370 had started his career trying to improve the sound of 0:08:40.410 --> 0:08:44.970 telephone conversations. Now he was leading a consortium of companies 0:08:44.970 --> 0:08:50.730 from several countries trying to make their CODEC the Chosen One. CODEC, 0:08:50.850 --> 0:08:55.050 by the way, is short for CODA decoder. These days, 0:08:55.050 --> 0:08:57.930 a CODEC is just a piece of software. Back in 0:08:58.010 --> 0:09:01.010 nineteen eighty eight, most people thought the winning audio compression 0:09:01.050 --> 0:09:03.690 CODEC would end up as some kind of box, a 0:09:03.730 --> 0:09:06.450 piece of electronic equipment that would be bought by labs 0:09:06.490 --> 0:09:11.410 and recording studios. I'm a fan of people like Karl Heinz, 0:09:12.210 --> 0:09:14.930 engineers who can work through problems and build things to 0:09:15.010 --> 0:09:17.970 solve them. This is partly because I'm not one of 0:09:18.010 --> 0:09:21.050 life's standardizers. I like to make things up as I 0:09:21.090 --> 0:09:24.010 go along. When I'm cooking, I'm the kind of cook 0:09:24.090 --> 0:09:27.370 who throws stuff in the pan eyeballing quantities. You know 0:09:27.370 --> 0:09:31.210 what I'm saying, yet without technical standards, most of the 0:09:31.290 --> 0:09:34.850 things I do other than cooking, I couldn't do. I 0:09:34.850 --> 0:09:38.690 couldn't record this podcast. You couldn't download it or stream 0:09:38.690 --> 0:09:42.450 it over the internet. The internet only works. The Internet 0:09:42.490 --> 0:09:46.370 could only work because everyone agrees to do something very 0:09:46.370 --> 0:09:50.210 complicated in exactly the same way as a species. We 0:09:50.250 --> 0:09:53.330 can't agree about anything, and yet we've agreed on that. 0:09:54.330 --> 0:10:00.530 It's incredible. The way humans perceive sound is also very complicated, 0:10:01.130 --> 0:10:04.530 as Karl heines Well knows. Would you explain a little 0:10:04.570 --> 0:10:08.170 about the mechanics of how the ear registers sound and 0:10:08.210 --> 0:10:10.850 transmits it to the brain yep. Of course there are 0:10:10.970 --> 0:10:15.610 short and long versions of set explanations. Maybe a quite 0:10:15.650 --> 0:10:23.250 short version please, okay. So, as you know, sound enters 0:10:23.250 --> 0:10:26.570 the pinner and the ear canal to the ear drum. 0:10:28.050 --> 0:10:32.210 Then we got the so called middle ears, ear drums, 0:10:32.250 --> 0:10:37.970 the arsicles, and then to the inner ear and on. 0:10:38.410 --> 0:10:41.210 In the inner ear we have the so called cochlera, 0:10:42.770 --> 0:10:45.290 which is a little thing. It is actually quite a 0:10:45.330 --> 0:10:49.170 long explanation. Karl Heinz goes on to talk about how 0:10:49.210 --> 0:10:53.050 tiny hairs in the cochlear are connected to neurons, which 0:10:53.130 --> 0:10:57.730 detects sound. Waves and send signals to the brain. This structure, 0:10:58.170 --> 0:11:02.170 which changes sound into electrical energy, goes by the name 0:11:02.610 --> 0:11:06.450 the organ of Corty. This would also be my band name, 0:11:06.890 --> 0:11:09.530 were it not already taken by a trio who describe 0:11:09.570 --> 0:11:12.890 themselves on bandcamp dot com as blues rock with an 0:11:12.930 --> 0:11:17.930 experimental edge. The actual organ of Corti in our heads 0:11:18.210 --> 0:11:22.570 functions as a kind of spectrum analyzer, separating sounds into 0:11:22.610 --> 0:11:26.650 different frequencies and telling the brain how strong each one is. 0:11:27.410 --> 0:11:31.010 So when you're trying to compress audio, when you're trying 0:11:31.010 --> 0:11:36.290 to work out what to keep and what can be removed, 0:11:36.850 --> 0:11:42.650 what you're trying to do is to use these limitations 0:11:42.650 --> 0:11:45.690 of the human ear to work out what parts of 0:11:45.690 --> 0:11:49.730 the signal can be removed. Is that correct? That's correct? 0:11:50.010 --> 0:11:53.530 In fact, a very simple way to phase it is 0:11:53.570 --> 0:11:59.450 that we try to store only that amount of information 0:12:00.050 --> 0:12:03.330 which goes from the inner ear to the brain. Karl 0:12:03.410 --> 0:12:06.730 Heinz explains that the organ of Corti can only handle 0:12:06.850 --> 0:12:10.890 so much information. Not every frequency that comes into the 0:12:10.970 --> 0:12:14.090 ear gets perceived by the brain. If there's an additional 0:12:14.210 --> 0:12:18.850 sound a's the same frequency, it doesn't change the pattern 0:12:18.890 --> 0:12:23.570 of moving hairselfs anymore, it doesn't change as a nurine firing. 0:12:24.290 --> 0:12:28.530 So the other signal, which is fainter, gets masked as 0:12:28.570 --> 0:12:32.050 we call it, right, So this is this is masking 0:12:32.490 --> 0:12:35.770 one of the key things that you're looking for when 0:12:36.490 --> 0:12:39.210 you're you're trying to work out what to keep and 0:12:39.250 --> 0:12:44.890 what to throw away. Exactly what to keep and what 0:12:45.010 --> 0:12:49.370 to throw away? This was the great technical question facing 0:12:49.370 --> 0:12:53.450 the audio engineers in Hanover in nineteen eighty eight. What 0:12:53.570 --> 0:12:56.610 they wanted to build was the very best and most 0:12:56.650 --> 0:13:02.010 efficient way to encode sound, one format used by everyone. 0:13:03.290 --> 0:13:06.450 It had to sound good, obviously, but it also had 0:13:06.490 --> 0:13:11.450 to be small because, just like our brains, phone lines 0:13:11.530 --> 0:13:14.770 or radio waves can only handle so much sound information 0:13:14.810 --> 0:13:19.010 at once. Therefore, it's good to throw away as much 0:13:19.050 --> 0:13:22.850 as possible. If one sound is masked by another, great 0:13:23.210 --> 0:13:29.730 just get rid of it. Being engineers, the engineers in 0:13:29.850 --> 0:13:34.290 Hanover had a very complicated formula for testing the competing proposals. 0:13:35.170 --> 0:13:38.410 The winner would need to be efficient and small, it 0:13:38.450 --> 0:13:41.010 would need to be simple to encode things and play 0:13:41.050 --> 0:13:43.970 them back, and of course it would have to sound good. 0:13:45.290 --> 0:13:48.690 But what does sound good mean? The engineers weren't just 0:13:48.770 --> 0:13:50.850 reading off numbers on a chart to find out what 0:13:50.930 --> 0:13:56.370 sounded best. They compare different compression algorithms by listening souls 0:13:56.410 --> 0:14:01.890 of final judgment was always listening tests. So what were 0:14:01.930 --> 0:14:07.050 the testers listening to? For quite some time as the 0:14:07.810 --> 0:14:11.810 task was to find difficult pieces of music because the 0:14:12.010 --> 0:14:17.210 very first attempts to do music compression already worked quite 0:14:17.290 --> 0:14:22.090 nicely for some signals and sounded horrible for us us. 0:14:22.570 --> 0:14:27.570 I had Glockenspiel as one early example of sounding horrible. 0:14:27.850 --> 0:14:32.410 What was easy easy was dance music, like if you 0:14:32.490 --> 0:14:37.410 had a full symphony orchestra or pop group which lots 0:14:37.410 --> 0:14:40.690 of sounds, that was easier. So it's when sounds are 0:14:40.810 --> 0:14:43.930 very distinct that it's easy to hear the limits of 0:14:44.010 --> 0:14:47.490 the compression. Is that correct? Correct? And one of the 0:14:47.530 --> 0:14:54.570 most difficult was Susan Vega's voice in her Solitude Standing 0:14:54.890 --> 0:15:06.690 appena piece Tom's Diner. It's a track on Suzanne vegas 0:15:06.690 --> 0:15:10.170 album Solitude Standing. Back in nineteen eighty seven. When the 0:15:10.210 --> 0:15:12.770 album was released, it became the epitome of a certain 0:15:12.810 --> 0:15:18.090 sort of unthreatening, good taste, yuppie dinner party music. Solitude 0:15:18.170 --> 0:15:20.170 Standing was the kind of record that people with new 0:15:20.210 --> 0:15:22.970 CD players would buy to showcase the performance of their 0:15:23.050 --> 0:15:26.770 high fire systems. This album went platinum in the US, 0:15:26.810 --> 0:15:29.850 and it was even more popular in Europe, so it's 0:15:29.850 --> 0:15:33.170 not surprising that a CD of Solitude Standing was floating 0:15:33.210 --> 0:15:36.730 around in some European lab where engineers were obsessing over 0:15:36.890 --> 0:15:45.450 digital audio standards. So Carl Hines and his colleagues compressed 0:15:45.450 --> 0:15:49.050 and listened to a lot of Suzanne Vega. They also 0:15:49.090 --> 0:15:53.850 listened to Tracy Chapman, Haydn, and Ornette Coleman. He might 0:15:53.890 --> 0:15:56.330 ask why they didn't test their system on even more 0:15:56.410 --> 0:16:00.570 varieties of music. Well, because it took eight hours to 0:16:00.690 --> 0:16:05.090 digitize just twenty seconds of sound. The song Tom's Diner 0:16:05.170 --> 0:16:08.170 alone is two minutes and nine seconds long. That would 0:16:08.170 --> 0:16:11.610 take more than two days of processing. So they used 0:16:11.650 --> 0:16:16.370 tiny snippets twenty seconds long and played them again and 0:16:16.530 --> 0:16:22.250 again and again. We tried it and it sounded horrible, 0:16:24.570 --> 0:16:29.290 really horrible. I am sitting in the morning at the 0:16:29.490 --> 0:16:33.170 diner on the corner, and this simple sound of a 0:16:33.250 --> 0:16:36.130 woman's voice turned out to be one of the hardest 0:16:36.170 --> 0:16:40.090 to get right. The problem was worst with Vega's vowels. 0:16:40.730 --> 0:16:44.610 When she sang, she distributed sound energy all across the 0:16:44.650 --> 0:16:47.970 audio spectrum. The algorithm was designed to make a kind 0:16:47.970 --> 0:16:50.210 of trade off. It could get one aspect of the 0:16:50.250 --> 0:16:53.930 sound perfect, but another would be sketchy. Finally they worked 0:16:53.970 --> 0:16:57.450 out a fix and those Doo doodoo started to sound 0:16:57.490 --> 0:17:01.250 pretty good. I am sitting in the morning at the 0:17:01.490 --> 0:17:05.290 diner on the corner, I am waiting at the counter. 0:17:05.650 --> 0:17:08.050 The problem was when the scores came in at a 0:17:08.130 --> 0:17:11.490 later meeting, no one could actually decide which of the 0:17:11.530 --> 0:17:15.730 fourteen competitors sounded best. The story of how it exactly 0:17:15.770 --> 0:17:18.890 went down is byzantine and technical. All you really need 0:17:18.890 --> 0:17:21.770 to know is that the process got bad tempered, and 0:17:21.850 --> 0:17:25.330 finally there was a compromise. At the hand of a conference, 0:17:25.370 --> 0:17:27.970 it had been decided the audio standard would have three 0:17:28.090 --> 0:17:32.330 different layers. Not three different standards, Oh no, not three 0:17:32.370 --> 0:17:36.530 different things. Layers like a cake, a nice tasty cake. 0:17:37.210 --> 0:17:39.930 And one of the three winners was the version developed 0:17:39.930 --> 0:17:43.850 by Karl Heinz. Its official name was Motion Picture Expert 0:17:43.890 --> 0:17:49.450 Group one Audio Layer Number three, or as it came 0:17:49.490 --> 0:17:53.490 to be known, the MP three, and the music industry 0:17:54.090 --> 0:18:08.930 would never be the same as Karl Heinz Brandenburg shook 0:18:09.170 --> 0:18:12.530 ends and was congratulated for his work on audio compression 0:18:13.370 --> 0:18:17.170 he had no idea what he had unleashed. He and 0:18:17.250 --> 0:18:21.370 his employer, a research lab called Fraunhofer, just wanted to 0:18:21.410 --> 0:18:24.530 sell some boxes of high end electronics to recording studios. 0:18:24.930 --> 0:18:29.890 We sought physical boxes and we licensed to companies how 0:18:29.970 --> 0:18:35.370 to build these physical boxes. But this was for professional use. 0:18:37.930 --> 0:18:44.690 We had software out there. And then some young guy 0:18:44.970 --> 0:18:50.650 from Australia used a Stone credit card number to buy 0:18:50.850 --> 0:18:55.090 one of these professional encoders. So he just built his 0:18:55.290 --> 0:19:01.770 own user interface, put our encoding library into it, put 0:19:01.810 --> 0:19:08.130 everything to GASA, and he told everybody to take off 0:19:08.290 --> 0:19:13.250 the software off their FTP sides. But of course knows 0:19:13.330 --> 0:19:17.770 this has been the same until now. In the Internet, 0:19:17.850 --> 0:19:27.530 you can't get things removed. It's still somewhere by now. 0:19:27.890 --> 0:19:31.610 It was nineteen ninety seven. The hacker eventually got arrested 0:19:32.450 --> 0:19:36.970 and the MP three eight the music industry. Within a 0:19:36.970 --> 0:19:40.850 few years, it had almost destroyed the compact disc. It 0:19:40.890 --> 0:19:44.450 cannibalized record company profits and made it harder for musicians 0:19:44.490 --> 0:19:49.130 to earn a living. In nineteen ninety nine, MP three 0:19:49.290 --> 0:19:52.650 surpassed sex as the most searched word on the Internet. 0:19:53.530 --> 0:19:55.970 By the late nineties, music was being pumped into our 0:19:56.010 --> 0:19:59.330 brains like a fire hose. Music fans like me could 0:19:59.370 --> 0:20:03.970 collect more than we could ever listen to. So much music. 0:20:05.570 --> 0:20:09.570 Of course, the technology was not without its critics. Do 0:20:09.650 --> 0:20:15.130 you know who you are? Metallica? I love and everything 0:20:15.170 --> 0:20:17.970 you do, except for that bad show your host of. 0:20:18.050 --> 0:20:19.810 You know what, maybe I wouldn't have to horn myself 0:20:19.850 --> 0:20:22.930 out if your kids didn't steal my music. So that's 0:20:23.010 --> 0:20:26.490 Marlon Wayne's and Lars Ulric doing a skit at the 0:20:26.570 --> 0:20:30.450 MTV Video Music Awards in two thousand. To me, the 0:20:30.490 --> 0:20:33.210 most interesting thing about the creation of the MP three 0:20:33.370 --> 0:20:36.370 isn't the story of Napster or the iPod or whatever. 0:20:37.010 --> 0:20:40.530 It's about what got kept and what got thrown away. 0:20:41.690 --> 0:20:44.090 If you're compressing a song down to ten percent of 0:20:44.130 --> 0:20:47.250 its original size, then what do you get rid of? 0:20:48.490 --> 0:20:52.410 Karl Heinz's MP three algorithm separates sound into two parts, 0:20:52.690 --> 0:20:55.490 the wheat and the chaff. The idea that such a 0:20:55.530 --> 0:21:00.170 decision could be subjective is displeasing to him, But I 0:21:00.210 --> 0:21:04.010 wonder can it really be objective? How do you tell 0:21:04.210 --> 0:21:10.130 information from noise? And what exactly is information? Is there 0:21:10.330 --> 0:21:13.170 like a living room or somewhere a little bit more 0:21:13.210 --> 0:21:16.130 away from drill we could probably do with the marking, 0:21:16.130 --> 0:21:20.490 and I'm really sorry about Let's try. There's a room. 0:21:20.810 --> 0:21:26.130 Yeah down here, let's see if that's the sound of me, 0:21:26.330 --> 0:21:30.090 My producers Hunter and Rider, and the science writer James Glick, 0:21:30.530 --> 0:21:37.930 trying to find somewhere quiet to talk in his apartment slightly. 0:21:39.570 --> 0:21:42.570 In nineteen eighty seven, James wrote a book called Chaos, 0:21:43.050 --> 0:21:46.410 all about the mathematical discovery of so called chaos theory. 0:21:47.570 --> 0:21:50.050 Oddly for a book about math, Chaos was something of 0:21:50.090 --> 0:21:54.330 a countercultural classic in nineties Britain. People skinned up joints 0:21:54.330 --> 0:21:58.210 on its cover, sitting under psychedelic posters of Mandelbrot sets 0:21:58.250 --> 0:22:01.690 and talking about the butterfly effect before their pills came 0:22:01.730 --> 0:22:03.930 on and they headed out into the fields to make 0:22:03.970 --> 0:22:06.930 crop circles. I don't know if James knows about his 0:22:06.970 --> 0:22:09.810 impact on the rave Ea and anyway, it's not really 0:22:09.850 --> 0:22:13.090 what I've come to talk to him about. It's louder here, 0:22:13.290 --> 0:22:15.570 it feels happy. Yeah, let's just go back to the 0:22:15.650 --> 0:22:18.450 kitchen table and we'll work with what we got. Someone 0:22:18.850 --> 0:22:22.650 somewhere in James's building is renovating. He says it could 0:22:22.650 --> 0:22:25.210 be several flaws away, but because of the acoustics. The 0:22:25.290 --> 0:22:29.370 sound of power tools is audible wherever we go. It 0:22:29.450 --> 0:22:31.610 is I suppose a sort of cosmic joke that we 0:22:31.690 --> 0:22:35.210 can't talk without noise getting onto the recording, because I've 0:22:35.250 --> 0:22:38.610 come to talk to him about noise, or rather about 0:22:38.610 --> 0:22:43.690 the difference between signal a noise. This binary opposition is 0:22:43.730 --> 0:22:47.250 at the core of the modern idea of information, and 0:22:47.410 --> 0:22:50.570 James has written a book about it, called The Information 0:22:51.090 --> 0:22:57.410 A History a Theory Flood. It's obviously an old word information, 0:22:57.490 --> 0:22:59.210 and you can look it up in the OED and 0:22:59.210 --> 0:23:02.490 it goes back many centuries, but it wasn't the kind 0:23:02.530 --> 0:23:04.690 of word it is now. It wasn't a common word, 0:23:04.690 --> 0:23:07.810 and it didn't refer to the stuff that we think 0:23:07.850 --> 0:23:14.410 of as information, the whole collection of sounds, words, images, 0:23:15.570 --> 0:23:20.210 types of knowledge that are encoded biologically. We understand information 0:23:20.250 --> 0:23:25.370 as a gigantic and extremely important category of stuff for us, 0:23:25.690 --> 0:23:30.850 and that began, I argue at a very specific moment 0:23:30.890 --> 0:23:33.530 of time. On the notes I wrote to prepare for 0:23:33.570 --> 0:23:38.410 this interview, one name appears again and again. Claude Shannon. 0:23:39.090 --> 0:23:41.810 He was a scientist who first introduced the idea that 0:23:41.850 --> 0:23:46.010 information was something you could measure. He published his Mathematical 0:23:46.050 --> 0:23:50.010 Theory of communication in nineteen forty eight, the first time 0:23:50.050 --> 0:23:54.330 he defines information the way a scientist needs to define 0:23:54.370 --> 0:23:56.890 it as a thing that you can measure. There's a 0:23:57.010 --> 0:23:59.210 unit of measure for information, and we all know what 0:23:59.250 --> 0:24:02.490 that is. It's the bit yes or no, on or off, 0:24:03.170 --> 0:24:07.610 zero or one, a very precise thing. Before that, the 0:24:07.690 --> 0:24:11.370 idea of measuring information didn't make any sense. It was like, 0:24:11.810 --> 0:24:16.250 could you have a scientific unit of measure for happiness 0:24:16.370 --> 0:24:21.890 or for anxiety or for fog? And then once you've 0:24:21.890 --> 0:24:25.730 got them in bits, you have something that applies to 0:24:25.770 --> 0:24:28.250 everything else. You can start to say how many bits 0:24:28.290 --> 0:24:31.250 are there in a book? How many bits of information 0:24:31.330 --> 0:24:33.650 are there in that painting? Does that? Is that a 0:24:33.730 --> 0:24:36.930 question that makes sense? As a writer, I think of 0:24:36.970 --> 0:24:40.610 myself as someone who deals with information that when I 0:24:40.730 --> 0:24:43.890 use that word, I'm not speaking in the mathematical terms 0:24:43.890 --> 0:24:48.570 of Claude Shannon. I'm someone who deals with meaning. I'm 0:24:48.570 --> 0:24:51.370 trying to make this podcast as meaningful as I can. 0:24:52.370 --> 0:24:54.930 When I do that, I think of it as being 0:24:55.050 --> 0:24:59.970 rich in information. But for Claude Shannon, it wasn't like 0:25:00.050 --> 0:25:04.970 that at all. Information had nothing to do with meaning. 0:25:05.410 --> 0:25:07.930 It is counterintuitive. It seems to us that if we 0:25:07.970 --> 0:25:10.890 start to talk about informa nation, apart from meaning you're 0:25:11.730 --> 0:25:16.170 you're talking about noise again. And Shannon's contemporaries also found 0:25:16.370 --> 0:25:19.570 this difficult to grasp, so he had to really emphasize that. 0:25:19.650 --> 0:25:22.770 He had to say again and again that the content 0:25:22.850 --> 0:25:26.210 of a message was irrelevant to the kind of mathematical 0:25:26.290 --> 0:25:30.370 treatment he was creating. He was measuring the bit content 0:25:30.610 --> 0:25:35.290 of a given message, and that bit content didn't depend 0:25:35.490 --> 0:25:39.810 on whether the message was crucially important or whether it 0:25:39.930 --> 0:25:42.970 was just a lot of nonsense. Can you just say 0:25:43.050 --> 0:25:45.330 the thing about this thing? There is my sentence about 0:25:45.330 --> 0:25:47.210 the signal and information again, and there's a bus that 0:25:47.330 --> 0:25:51.410 drove by. Oh, but what was it the signal? Give 0:25:51.490 --> 0:25:57.450 me a little more these interruptions demonstrating what we're talking about. 0:25:58.730 --> 0:26:03.370 Is the bus passing outside signal or noise? Could it 0:26:03.410 --> 0:26:07.090 be both? I found myself when I was thinking about 0:26:07.090 --> 0:26:11.290 this question working on the book, thinking about particular pieces 0:26:11.330 --> 0:26:19.650 of music like a Bach prelude. I mean, I'm thinking 0:26:19.690 --> 0:26:23.370 particularly of the first one in C major, yeah, which 0:26:23.770 --> 0:26:28.050 you know is on the Voyager Golden Records that's making 0:26:28.050 --> 0:26:34.970 its way to to other galaxies. Because Carl Sagan and 0:26:35.010 --> 0:26:37.450 his committee thought it would reveal something about us as 0:26:37.530 --> 0:26:42.490 human beings. Well this, you ask yourself, how much information 0:26:42.850 --> 0:26:52.530 is in this bach prelude? You play it, and the 0:26:52.570 --> 0:26:55.130 first thing you notice is it's very repetitious. It's the 0:26:55.250 --> 0:26:58.290 same thing over and over again, except that it keeps 0:26:58.410 --> 0:27:01.210 changing a little bit. And if you were Claude Shannon, 0:27:01.610 --> 0:27:06.410 you would say, well, I could compress this algorithmically, because 0:27:06.490 --> 0:27:13.930 all I really need is the changes. The stuff that 0:27:14.130 --> 0:27:23.690 repeats itself is not new information. And as a listener, 0:27:24.530 --> 0:27:27.770 you also recognize the tension between these two things. You 0:27:27.850 --> 0:27:33.130 recognize that if it's the patterns that you sense that 0:27:33.170 --> 0:27:36.170 make it beautiful, and it's also the fact that the 0:27:36.250 --> 0:27:41.330 patterns do not repeat themselves perfectly that make it exquisitely beautiful. 0:27:47.690 --> 0:27:51.930 Claude Shannon's revelation was that information is a measure of complexity, 0:27:52.690 --> 0:27:56.850 so you can assign it a number. But being complex 0:27:57.130 --> 0:28:01.330 is not the same as being meaningful. Things can be 0:28:01.450 --> 0:28:05.730 complex and meaningless, as anyone alive in twenty twenty can attest. 0:28:09.330 --> 0:28:13.290 When it comes to sounding good, let alone esthetic beauty. 0:28:13.970 --> 0:28:19.130 It's not about information. Shannon leaves us cold. He leaves 0:28:19.170 --> 0:28:22.850 us empty for exactly that reason that he tried to 0:28:22.890 --> 0:28:27.570 remove the human element from the equation, he explicitly removed 0:28:27.650 --> 0:28:30.410 meaning from the equation. And meaning is what we want. 0:28:30.650 --> 0:28:33.930 Meaning is what we are desperate for, So how do 0:28:34.010 --> 0:28:40.130 we find it? Meaning is what we're desperate for? And 0:28:40.290 --> 0:28:43.490 James Blick's main question is how do we find it? 0:28:45.050 --> 0:28:48.930 I've always thought of meaning as something you make, for example, 0:28:48.970 --> 0:28:54.690 through words and stories, but maybe meaning is now something 0:28:55.170 --> 0:29:22.650 we have to reclaim from our algorithms. The listening test 0:29:22.690 --> 0:29:26.890 for the MP three were done in Germany, and so 0:29:27.330 --> 0:29:34.290 they were predominantly engineers, and I think mostly white, uh, 0:29:34.810 --> 0:29:42.210 sort of German men, professional, sort of audio professionals. This 0:29:42.410 --> 0:29:47.250 is Ryan McGuire, a composer based in Chicago. It was 0:29:47.370 --> 0:29:50.010 his ghostly music that I found on the Internet a 0:29:50.010 --> 0:29:53.930 few years ago while I was writing my novel White Tears. 0:29:54.850 --> 0:29:57.490 His music is what led me down the rabbit hole 0:29:57.530 --> 0:30:04.010 of audio compression and the MP three. Ryan's obsessed with 0:30:04.130 --> 0:30:07.170 the early listening tests that led to the MP three, 0:30:07.810 --> 0:30:12.490 all those German engineer is with the CDs. I don't 0:30:12.530 --> 0:30:19.090 think it was particularly systematic, but the songs for the 0:30:19.210 --> 0:30:22.930 MP three listening tests, the early MP three listening tests, 0:30:23.330 --> 0:30:25.730 you can tell that they were trying to have a 0:30:25.770 --> 0:30:29.690