WEBVTT - The Machine Speaks 0:00:03.040 --> 0:00:07.000 Welcome to Stuff to Blow Your Mind, a production of iHeartRadio. 0:00:12.800 --> 0:00:14.520 Hey, welcome to Stuff to Blow Your Mind. 0:00:14.600 --> 0:00:18.000 My name is Robert Lamb and I am Joe McCormick, 0:00:18.040 --> 0:00:21.480 and today we're going to be talking about some early 0:00:22.120 --> 0:00:26.000 voice synthesis machines. Rob I actually got interested in this 0:00:26.120 --> 0:00:30.560 topic because last week, when we were watching the Weird 0:00:30.560 --> 0:00:33.680 House Cinema movie The Black Hole, I was thinking about 0:00:33.840 --> 0:00:36.879 Roddy McDowell's voice when he's doing a voice for the 0:00:36.960 --> 0:00:40.280 robot character who shares a lot of proverbs with the 0:00:40.360 --> 0:00:44.360 human characters. And I kept listening to his line delivery 0:00:44.400 --> 0:00:47.159 and I couldn't decide if he was trying to do 0:00:47.320 --> 0:00:50.360 quote robot voice or not. He seemed to kind of 0:00:50.360 --> 0:00:51.559 dip in and out of it. You know what I 0:00:51.600 --> 0:00:54.200 mean When I say robot voice, where a character's playing 0:00:54.240 --> 0:00:57.640 a robot and they say things like this, Well. 0:00:57.440 --> 0:00:59.600 Of course I know what you are talking about, Joe. 0:01:00.360 --> 0:01:03.760 I got kind of interested in the history of robot voice. 0:01:03.840 --> 0:01:06.440 I was like, where does that come from? And I 0:01:06.480 --> 0:01:08.160 was digging around a little. I'm sure there is a 0:01:08.160 --> 0:01:10.800 good answer on that, but I don't know. My short 0:01:10.800 --> 0:01:14.120 search didn't really turn up anything interesting, but it did 0:01:14.319 --> 0:01:18.160 lead me indirectly to what we're talking about today, which is, 0:01:18.280 --> 0:01:23.280 of course, we have the voice synthesis systems that are 0:01:23.400 --> 0:01:26.080 largely digital today. Before that you had a lot of 0:01:26.520 --> 0:01:32.319 electrical and electro mechanical systems for synthesizing human voices. But 0:01:32.400 --> 0:01:35.039 actually there is an even earlier generation, which are the 0:01:35.200 --> 0:01:40.600 purely mechanical voice synthesizers before electricity even came into the picture. 0:01:40.959 --> 0:01:44.440 And that is what really stole my heart, especially one 0:01:44.520 --> 0:01:46.760 particular machine of this type that I'm going to talk 0:01:46.760 --> 0:01:48.560 about in the second half of this episode. 0:01:48.560 --> 0:01:51.720 I think, yeah, this is a fascinating topic, in part 0:01:51.800 --> 0:01:54.280 because look at it. Look at where we are now, right, 0:01:54.320 --> 0:01:57.360 it's easy today in our Internet age for just the 0:01:57.400 --> 0:02:01.120 average Internet user to engage with their as chat bots 0:02:01.160 --> 0:02:05.920 and generative AI, text to speech and so forth. And 0:02:05.960 --> 0:02:09.640 you know, so we're able to interact with an artifact, 0:02:09.680 --> 0:02:12.400 a thing that reflects human will, that has been designed 0:02:12.440 --> 0:02:15.440 to do key and telling things that have long been 0:02:15.480 --> 0:02:19.680 the hallmarks of human activity, artistic generation, creative writing and 0:02:19.840 --> 0:02:24.600 conversation or especially speech. And of course it's you know, 0:02:24.600 --> 0:02:27.520 it's easy nowadays to do that, right, to transform into 0:02:27.680 --> 0:02:31.640 audible or even video content. What is either written by 0:02:31.720 --> 0:02:34.520 human or or created with some sort of a chat 0:02:34.720 --> 0:02:39.160 bot machine, and the results may be amusing, that may 0:02:39.480 --> 0:02:42.280 may be disastrous. But we're in this age where the 0:02:42.320 --> 0:02:45.240 idea of the machine speaking is not in and of 0:02:45.280 --> 0:02:49.240 itself groundbreaking, or at least if it is groundbreaking, or 0:02:49.240 --> 0:02:53.600 if it's amazing, it's it's that it's a lower level 0:02:53.600 --> 0:02:55.720 of amazement compared to previous ages. 0:02:56.320 --> 0:02:59.040 Well, as you say, it's very integrated into modern technology. 0:02:59.080 --> 0:03:02.040 So there's you know, Series and Alexa, all these like 0:03:02.240 --> 0:03:06.080 home devices that speak, GPS devices for the car, you know, 0:03:06.160 --> 0:03:09.040 that speak to you, but almost all of them are 0:03:09.080 --> 0:03:11.880 still the subject of amusement if you actually pay attention 0:03:11.919 --> 0:03:14.880 to what the voice sounds like, you know, like reading 0:03:14.960 --> 0:03:17.400 emotions into the voice that's telling you what to do 0:03:17.440 --> 0:03:20.480 as you're driving. That always makes me laugh because it 0:03:20.520 --> 0:03:21.960 always seems a little bit annoyed. 0:03:22.680 --> 0:03:25.919 Yeah, yeah, what's Series whole deal that sort of thing, right? 0:03:27.760 --> 0:03:30.280 You know. The other interesting angle on all this is 0:03:30.320 --> 0:03:35.560 that are modern technological advancements here, or even some of 0:03:35.600 --> 0:03:38.720 the historic technological advancements like they are kind of the 0:03:38.760 --> 0:03:42.960 echo of a more ancient longing for this sort of thing. 0:03:43.720 --> 0:03:47.880 It connects to something that's just fascinated us for a 0:03:47.920 --> 0:03:52.440 long time, the idea generally of non human entities engaging 0:03:52.480 --> 0:03:55.040 in speech, and you could you could do absolutely wild 0:03:55.120 --> 0:03:59.280 chasing down the various divisions of this right, the various myths, legends, 0:03:59.280 --> 0:04:04.920 and traditions concerning the speech of animals, plants, inorganic materials, 0:04:04.960 --> 0:04:10.800 supernatural entities. You know, voices seemingly internal but also external 0:04:11.000 --> 0:04:12.520 to our individual experience. 0:04:12.960 --> 0:04:15.120 Though I would say there is an interesting thing about 0:04:15.360 --> 0:04:19.680 machines or human or automata or human artifacts in general, 0:04:20.000 --> 0:04:24.520 when compared to imagining an animal speak or any other 0:04:24.839 --> 0:04:27.680 usually not speaking things starting to speak, which is, if 0:04:27.680 --> 0:04:30.360 you're talking about a machine that does it, that means 0:04:30.520 --> 0:04:33.440 somebody has to make that machine, and somebody has to 0:04:33.480 --> 0:04:36.120 work that machine. And it kind of reminds me of 0:04:36.240 --> 0:04:39.520 the idea of grammar in language. You know. The interesting 0:04:39.560 --> 0:04:42.640 thing about grammar is that when we use language, we 0:04:42.680 --> 0:04:45.880 all use grammar, so we have an intuitive grasp of 0:04:45.920 --> 0:04:49.320 the rules of grammar. But without serious study, people can't 0:04:49.320 --> 0:04:51.560 actually tell you what those rules are. And so, like 0:04:51.600 --> 0:04:53.599 you know, that had to be in a sense a 0:04:53.880 --> 0:04:57.479 science to back engineer the rules of grammar that we 0:04:57.600 --> 0:05:01.280 use intuitively to like make them systematic and you know, 0:05:01.360 --> 0:05:04.599 actually discover what those rules are the same thing could 0:05:04.600 --> 0:05:08.480 be said about the phonetic rules that produce the intelligible speech. 0:05:08.839 --> 0:05:11.400 We can all do it if we can speak, but 0:05:11.680 --> 0:05:16.720 we don't necessarily understand what the individual physical properties of 0:05:16.800 --> 0:05:19.800 a word are, and so we wouldn't necessarily know how 0:05:19.839 --> 0:05:22.359 to make that same word come out of a machine. 0:05:23.160 --> 0:05:25.120 Yeah, there are all these things that you have to 0:05:25.160 --> 0:05:28.800 deconstruct before you can attempt to reproduce it artificially. And 0:05:28.839 --> 0:05:32.000 we see that time and time again with in robotics, 0:05:32.000 --> 0:05:35.560 for example. You know, things that we take for granted 0:05:35.880 --> 0:05:39.600 concerning human movement, just about anything else you can imagine, 0:05:39.640 --> 0:05:42.200 it becomes so much more difficult to try and reproduce 0:05:42.200 --> 0:05:45.039 that you've got to understand what it actually is on 0:05:45.080 --> 0:05:46.599 an entirely new level. First. 0:05:47.279 --> 0:05:50.279 Now, I am to understand that before anybody actually made 0:05:50.320 --> 0:05:53.719 a machine that could approximate or synthesize a human voice 0:05:53.760 --> 0:05:57.320 and produce intelligible speech, people were thinking about this as 0:05:57.320 --> 0:05:58.080 a concept. 0:05:58.960 --> 0:06:01.920 Yeah, yeah, this is not surprising, you know, this is uh, 0:06:01.960 --> 0:06:05.520 this is kind of the the the meat of science fiction. 0:06:05.680 --> 0:06:07.400 Right before we can do it, we dream of it. 0:06:07.960 --> 0:06:11.680 One way or another, no matter what our exact grasp 0:06:11.720 --> 0:06:14.600 of science happens to be. It always reminds me of 0:06:14.640 --> 0:06:19.840 that line in William Gibson's Neuromancer where the character has 0:06:19.880 --> 0:06:22.880 made a deal, a pact with a powerful ai and 0:06:22.920 --> 0:06:24.960 it's pointed out like, this is the sort of thing 0:06:25.040 --> 0:06:28.520 that in you know, centuries ago, people only dreamed of 0:06:28.640 --> 0:06:31.520 making a deal with a devil, and now we've made 0:06:31.560 --> 0:06:37.840 it possible through our ingenuity and invention. Congratulations, yeah, so 0:06:37.839 --> 0:06:41.680 so yeah. Narrowing down here into generally the realm of 0:06:43.360 --> 0:06:49.919 alleged human creations that through at least partial technology, but 0:06:50.080 --> 0:06:53.520 also sometimes wizardry and alchemy and other things that are 0:06:53.600 --> 0:06:56.520 kind of like you know, bunched in there together with 0:06:56.520 --> 0:07:00.880 with actual technology to create some sort of a device 0:07:01.000 --> 0:07:03.440 capable of speech. And then there are some also some 0:07:03.520 --> 0:07:05.719 related things that are tied in there as well, and 0:07:05.800 --> 0:07:08.040 a lot of it comes down to the idea of 0:07:08.440 --> 0:07:12.440 a head, an artificial head that speaks. 0:07:13.080 --> 0:07:16.400 I found something so loaded and revealing about that. As 0:07:16.400 --> 0:07:19.360 a fact, the history of these machines, so many of 0:07:19.400 --> 0:07:23.600 them had, whether real or imagined, these machines, so many 0:07:23.640 --> 0:07:27.480 of them early on had heads or faces, so like 0:07:27.520 --> 0:07:30.000 it wouldn't just be a speaker like you would have today. 0:07:30.080 --> 0:07:33.080 That's you know, it's just a mechanical device for making 0:07:33.120 --> 0:07:36.120 the sound. It's like that the presence of a head 0:07:36.240 --> 0:07:39.840 or a face was considered important or at least desirable. 0:07:40.600 --> 0:07:42.440 Yeah, and I wondered to what extent part of it 0:07:42.480 --> 0:07:46.560 is just an echo of these earlier ideas. So going 0:07:46.720 --> 0:07:50.280 to run through a few of these here. One of 0:07:50.320 --> 0:07:53.920 the most famous, mainly from a literary tradition, as we'll 0:07:54.080 --> 0:07:57.560 discuss here, is the idea of the brazen head. And 0:07:58.000 --> 0:08:00.080 ultimately I guess there's more than one brazen head. We 0:08:00.120 --> 0:08:05.320 can say brazen heads artificial heads that could speak. There's 0:08:05.360 --> 0:08:08.960 a basically a lot of these stories concerned thirteenth century 0:08:09.000 --> 0:08:13.160 English philosopher and Franciscan friar Roger Bacon, who's come up 0:08:13.160 --> 0:08:16.920 on the show before, though this particular version of the 0:08:16.920 --> 0:08:20.000 story doesn't seem to emerge until the sixteenth century, and 0:08:20.080 --> 0:08:23.240 it does so within the works of contemporary drama. 0:08:23.760 --> 0:08:25.960 I think we talked about Roger Bacon at length in 0:08:26.000 --> 0:08:28.960 an episode we did about the invention of fireworks, which 0:08:29.560 --> 0:08:31.920 may come back and feature again in the feed soon. 0:08:32.600 --> 0:08:34.559 Yeah. Yeah, I believe you're right, Yeah, I think Bacon 0:08:34.559 --> 0:08:37.800 did come up in that he had a reputation as 0:08:37.840 --> 0:08:41.559 not only a very learned man in both natural philosophy 0:08:41.559 --> 0:08:44.920 and theology, and I should drive home definitely existed. I 0:08:44.920 --> 0:08:47.160 don't think there's any doubt that there was a Roger Bacon. 0:08:47.640 --> 0:08:49.880 But then there are all these other stories that he 0:08:50.000 --> 0:08:54.439 was also potentially a wizard who was capable of producing 0:08:54.559 --> 0:09:00.480 fabulous automata, either through amazing feats of clockwork engineuity that 0:09:00.520 --> 0:09:03.760 I think would many would say was ultimately, you know, 0:09:04.240 --> 0:09:08.840 impossible during his time period, or failing that he was 0:09:08.880 --> 0:09:11.920 into alchemy and of course dark dank necromancy. 0:09:12.520 --> 0:09:14.600 I think the way I conceive of Roger Bacon is 0:09:14.600 --> 0:09:17.920 that he of course was a real figure. He was 0:09:18.679 --> 0:09:23.080 of great intellectual note and significance, but much about his 0:09:23.200 --> 0:09:26.880 sort of general reputation is kind of legendary, if that 0:09:26.880 --> 0:09:28.400 makes sense. I mean, there are many things we know 0:09:28.480 --> 0:09:32.160 about him that are true, but there's also just sort 0:09:32.160 --> 0:09:34.720 of an aura or a vibe about him that is 0:09:34.760 --> 0:09:36.160 not really based in reality. 0:09:36.840 --> 0:09:39.920 Yeah, I mean, he becomes a character in literature, especially 0:09:39.960 --> 0:09:42.800 in these accounts, so you can sort of look at 0:09:42.800 --> 0:09:46.199 the different phases like historic individual ideas and you know, 0:09:46.760 --> 0:09:51.040 misunderstandings have said real life individual and then eventually that 0:09:51.080 --> 0:09:53.920 echoes into the fictional version of the person. 0:09:53.920 --> 0:09:55.960 Which that's the more like wizard version. 0:09:56.720 --> 0:09:59.400 Yeah, yeah, And so there are a few different examples 0:09:59.400 --> 0:10:01.760 of this. This is like a popular motif for a while. 0:10:02.160 --> 0:10:06.040 There's a sixteenth century prose romance titled the Famous History 0:10:06.040 --> 0:10:08.520 of Friar Bacon, and it tells of Bacon trying to 0:10:08.960 --> 0:10:11.480 give a replica of a human head speech and having 0:10:11.520 --> 0:10:15.360 to call in the devil for help. Cool. Other versions 0:10:15.360 --> 0:10:17.520 of this tale describe it as an artificial head given 0:10:17.600 --> 0:10:21.480 life by demons, which was capable of spontaneous speech and 0:10:21.559 --> 0:10:23.199 of course telling the future. 0:10:24.600 --> 0:10:26.400 I mean, what else would you tell? 0:10:26.480 --> 0:10:31.400 Right right? Robert Greene's sixteen thirty play Friar Bacon and 0:10:31.440 --> 0:10:36.920 Friar Bungay mentions this several times, citing quote Bacon's necromantic 0:10:37.000 --> 0:10:41.880 skill and heads of Brass that quote can utter any voice. 0:10:42.200 --> 0:10:44.560 The idea that's exploring both of these works is that 0:10:44.600 --> 0:10:47.520 Bacon wished to build a wall of brass around Britain 0:10:47.800 --> 0:10:51.280 with the help of the Brazen Head. He fails and 0:10:51.320 --> 0:10:52.199 the head explodes. 0:10:54.640 --> 0:10:56.480 Why have I never heard this? Site? It as like 0:10:56.520 --> 0:10:58.199 an early science fiction tale. 0:10:58.679 --> 0:11:00.720 I don't know, I'm probably not doing do diligence on 0:11:00.840 --> 0:11:03.640 exactly what happens and everything at It's like saying, well, 0:11:03.640 --> 0:11:06.319 in Star Wars, the bad guys make one planet to 0:11:06.320 --> 0:11:08.520 blow up another planet, and then the planet they may 0:11:08.559 --> 0:11:11.360 blows up. You know, That's that's really skipping over a 0:11:11.360 --> 0:11:13.640 lot of the nuance. And so I think there's there's 0:11:13.679 --> 0:11:15.839 inevitably more nuance here, but I just I didn't get 0:11:15.840 --> 0:11:19.840 into it. Okay, So this idea of the satanic brasshead 0:11:19.880 --> 0:11:23.040 of Roger Bacon persists despite the fact that there's no 0:11:23.080 --> 0:11:27.040 indication that anything like this even created purely through technology 0:11:27.120 --> 0:11:30.760 and not satanic wizardry was part of Bacon's world. He 0:11:30.840 --> 0:11:35.320 was interested in optics and certainly various instruments scientific instruments 0:11:35.320 --> 0:11:37.240 of brass of the day, but there's no indication that 0:11:37.280 --> 0:11:39.679 he ever built an artificial head and tried to get 0:11:39.720 --> 0:11:40.320 it to speak. 0:11:40.640 --> 0:11:42.880 Okay, so this is part of the wizard aura, not 0:11:43.000 --> 0:11:44.920 part of his biography. 0:11:44.840 --> 0:11:47.600 Right though, you know, we have to drive home to 0:11:47.840 --> 0:11:51.200 is it possible that Roger Bacon as a hobby did 0:11:51.200 --> 0:11:53.120 what he could to create you know, I mean, it's 0:11:53.160 --> 0:11:55.760 possible it's not, you know, I don't think he would 0:11:55.760 --> 0:11:58.040 have gotten to speak. But there are various sort of 0:11:58.360 --> 0:12:00.880 ways you could interpret this as having some basis in 0:12:00.960 --> 0:12:06.040 reality that doesn't involve magic or super science of the day. Okay, now, 0:12:06.240 --> 0:12:08.600 I went to my bookshelf and I pulled off my 0:12:08.679 --> 0:12:12.240 dusty copy of Brewers Dictionary Phrase and fable and provides 0:12:12.280 --> 0:12:15.360 a little more insight on the legend quote. It was 0:12:15.360 --> 0:12:18.520 said if Bacon heard it speak, he would succeed in 0:12:18.559 --> 0:12:22.720 his projects, if not, he would fail. His familiar Miles 0:12:22.840 --> 0:12:25.560 was set to watch, and while Bacon slept, the head 0:12:25.640 --> 0:12:29.960 spoke thrice. Time is half an hour later, it said 0:12:30.320 --> 0:12:34.440 time was in another half hour, it said times past 0:12:34.880 --> 0:12:35.640 fell down. 0:12:35.440 --> 0:12:39.199 And was broken to atoms to atoms to atoms. Yes, 0:12:40.000 --> 0:12:43.720 surely Adams means something different here, Adams right when discovered 0:12:43.760 --> 0:12:46.200 at the time, I think it just means like small 0:12:46.240 --> 0:12:47.080 parts or something. 0:12:47.200 --> 0:12:49.839 Yeah, yes, yeah, that. 0:12:49.760 --> 0:12:52.559 Would be hilarious if it was literally broken to atoms. 0:12:53.160 --> 0:12:54.960 Yeah. So I don't know if it works, though. It 0:12:54.960 --> 0:12:57.280 sounds like it's kind of an alarm clock that explodes. 0:12:57.679 --> 0:13:01.160 Well, but I don't understand the difference between time was 0:13:01.360 --> 0:13:03.960 and time's passed. They're both past tens. 0:13:04.480 --> 0:13:08.280 Hmmm, that's a good point. Time is time was, Time's past. 0:13:08.360 --> 0:13:11.120 It seems like you would want the president there somewhere. 0:13:11.160 --> 0:13:15.920 But yeah, that's that's what it allegedly said. And you'll 0:13:15.920 --> 0:13:21.160 you'll find woodcuts that have this this motif on them 0:13:21.160 --> 0:13:23.000 as well. 0:13:22.200 --> 0:13:24.320 Like it would make more sense if it said the 0:13:24.360 --> 0:13:27.560 three things where time will be, time is, time was, 0:13:28.480 --> 0:13:31.000 but this seems more like time is, time was, time 0:13:31.240 --> 0:13:31.719 was was. 0:13:33.840 --> 0:13:36.880 Now. Brewers notes that reference to the references to the 0:13:36.920 --> 0:13:40.400 brazen head are just common in literature, appearing frequently in 0:13:40.480 --> 0:13:45.640 early romances but with Eastern origins, though it doesn't get 0:13:45.679 --> 0:13:48.720 into that a lot elsewhere in the volume. It's also 0:13:48.880 --> 0:13:54.600 noted that artificial heads that speak occur elsewhere as well. 0:13:54.640 --> 0:13:56.520 And some of these are brazen heads, and some of 0:13:56.520 --> 0:13:58.240 these are other things, but they're kind of I think 0:13:58.240 --> 0:14:01.040 it's important to run through brief some of these examples 0:14:01.040 --> 0:14:03.640 because they kind of paint a picture of not only 0:14:03.679 --> 0:14:07.160 some of these other ideas of artificial heads speaking and 0:14:07.200 --> 0:14:14.800 telling the future, but related non technological non artifacts that 0:14:15.200 --> 0:14:20.520 kind of help inform what we think technology can do. Okay, okay, 0:14:20.560 --> 0:14:22.520 so one of them is a brazen head in the 0:14:22.560 --> 0:14:25.520 possession of Pope Sylvester the Second in the tenth century, 0:14:25.800 --> 0:14:30.280 which he also constructed, and misinterpretations of its utterances could 0:14:30.320 --> 0:14:31.320 prove disastrous. 0:14:31.800 --> 0:14:35.840 Oh, is this also believed to be satanic in some way? 0:14:37.200 --> 0:14:42.480 I didn't go too deep on satanic implications, but possibly, I. 0:14:42.440 --> 0:14:44.800 Guess it would depend on if this legend is associated 0:14:44.840 --> 0:14:47.480 with pro Pope Sylvester or anti Pope Sylvester. 0:14:47.600 --> 0:14:51.040 Sources right, right, but you can definitely see that they're 0:14:51.160 --> 0:14:53.080 in the head itself, regardless of what's supposed to be 0:14:53.120 --> 0:14:56.520 powering it. Like this, it ties into two oracular traditions. 0:14:56.560 --> 0:14:59.360 You know, the idea that here is this thing that 0:14:59.440 --> 0:15:02.640 can give you you cryptic wisdom if you have the 0:15:02.680 --> 0:15:05.880 wisdom to decipher what it's telling you. Another example it's 0:15:05.880 --> 0:15:09.680 brought up in Brewers is or the Colossi of Memnon, 0:15:10.240 --> 0:15:12.280 which we did at least a whole I don't know. 0:15:12.280 --> 0:15:14.400 I can't remember as one episode or multiple episodes, but 0:15:14.440 --> 0:15:17.840 we discussed this on stuff to blow your mind. This 0:15:17.880 --> 0:15:19.960 is a fascinating topic in and of itself. 0:15:20.200 --> 0:15:22.560 This was basically, I think a statue or a pair 0:15:22.600 --> 0:15:25.680 of statues a part of sort of a ruins complex 0:15:25.840 --> 0:15:30.680 that was famous during in Roman Egypt as basically because 0:15:30.720 --> 0:15:33.640 it would make sounds, and there were different theories about 0:15:33.680 --> 0:15:35.160 how it made sounds and why. 0:15:35.480 --> 0:15:38.480 Yeah, yeah, it seems like I think some said it 0:15:38.520 --> 0:15:40.880 was capable of speech, but generally it's described as singing 0:15:41.040 --> 0:15:44.160 or some sort of a note. And as we discussed, 0:15:44.400 --> 0:15:47.400 while there are some I think unlikely theories regarding the 0:15:47.480 --> 0:15:51.280 use of some sort of intentional sound generating device or devices, 0:15:51.520 --> 0:15:53.920 it seems like a more likely explanation would have to 0:15:54.000 --> 0:15:57.440 do with peculiarities of the stone as it heated in 0:15:57.520 --> 0:16:00.480 the sun and then cooled at night. Anyway, go back 0:16:00.520 --> 0:16:01.760 and listen to that episode if you want to know 0:16:01.800 --> 0:16:04.160 about them. They have a pretty fascinating history. 0:16:04.360 --> 0:16:06.200 Will remember better in the original. 0:16:07.000 --> 0:16:12.080 Yes, there's the head of Orpheus at Lesbos, predicting the 0:16:12.080 --> 0:16:15.640 doom and death of Cyrus the Great. However, I believe 0:16:15.680 --> 0:16:17.960 this is generally thought to be the actual head of 0:16:18.000 --> 0:16:20.840 the hero Orpheus after he was torn apart by the 0:16:20.880 --> 0:16:25.080 main ads of Dionysus during a bacchanalia for the sin 0:16:25.160 --> 0:16:28.600 of worshiping Apollo or having worshiped Apollo. I'm not sure 0:16:28.600 --> 0:16:31.960 what the exact charge was, but still a prophetic disembodied 0:16:32.000 --> 0:16:35.760 head that still continues to speak. Brewers also mentions the 0:16:35.800 --> 0:16:40.000 head of Minos brought by Odin to Scandinavia, which I 0:16:40.040 --> 0:16:42.640 didn't know what to make of this, because Minos is, 0:16:42.680 --> 0:16:49.400 of course the mythical king of Crete, who we've discussed 0:16:49.400 --> 0:16:51.920 on the show before as well. I think the actual 0:16:52.040 --> 0:16:56.120 figure in reference here might be Nimir, the god of wisdom, 0:16:56.160 --> 0:17:01.720 that is beheaded in the Aservaniir War. Odin claims this 0:17:01.840 --> 0:17:05.000 head and it continues to speak secret wisdom. Again, this 0:17:05.080 --> 0:17:07.240 is another one that's not a mechanical head. It's the 0:17:07.240 --> 0:17:10.879 head of an actual defeated divine being that continues to 0:17:10.920 --> 0:17:15.040 live on and to speak. There are tales of Albertus 0:17:15.080 --> 0:17:18.800 Magnus having an earthen head which during the thirteenth century 0:17:18.960 --> 0:17:22.720 was said to speak and move until Thomas Aquinas breaks 0:17:22.720 --> 0:17:25.639 it by accident, and Magnus says, there goes the labor 0:17:25.680 --> 0:17:30.360 of thirty years because now it's broken. So I don't 0:17:30.359 --> 0:17:32.560 know what to make of that one either completely. But 0:17:32.600 --> 0:17:36.120 again we see this motif of a fabulous artificial head 0:17:36.119 --> 0:17:39.520 that speaks, that manages to break one way or another, 0:17:39.560 --> 0:17:43.840 either something fails, somebody knocks it over, or you know 0:17:43.920 --> 0:17:47.440 it explodes after you hit this neooze alarm twice. Then 0:17:47.440 --> 0:17:51.560 there's Alexander's statue of Ascalapius, the Greek god of medicine, 0:17:51.800 --> 0:17:55.440 that was said to speak, but Lucian wrote that the 0:17:55.480 --> 0:17:59.680 sounds came via a concealed man who spoke through tubes. 0:18:00.280 --> 0:18:05.120 So here's an example of some sort of of a creation. 0:18:06.359 --> 0:18:08.199 I guess it depends how you look. Either a statue 0:18:08.240 --> 0:18:13.480 that isn't intended to speak, or through supernatural machinations speaks, 0:18:13.640 --> 0:18:16.800 but according to Lucian, it's in neither of those. It's 0:18:16.840 --> 0:18:19.280 just tubes and some guy like hiding in the bushes 0:18:19.320 --> 0:18:22.560 speaking through the tubes, which is still clever and still technological, 0:18:23.320 --> 0:18:24.280 but is trickery. 0:18:24.920 --> 0:18:28.440 Nonetheless, I think the Lucian you're alluding to there is 0:18:28.520 --> 0:18:30.359 Luci of sam Masada, Is that right? 0:18:30.680 --> 0:18:31.119 I believe? 0:18:31.200 --> 0:18:31.240 So? 0:18:31.440 --> 0:18:34.240 Yes, yeah, this is like this was an ancient satirist 0:18:34.280 --> 0:18:38.879 from Syria who is quite hilarious and was kind of 0:18:38.920 --> 0:18:42.440 a skeptic debunker of the of like the second century CE, 0:18:42.600 --> 0:18:46.200 which is sort of strange, but he was in that 0:18:46.240 --> 0:18:50.480 mold and he made like vicious mockery of people of 0:18:50.520 --> 0:18:54.119 all sorts and different philosophies and stuff, and also wrote 0:18:54.160 --> 0:18:56.600 a satire that some people have considered one of the 0:18:56.640 --> 0:18:58.119 earliest forms of science fiction. 0:18:59.640 --> 0:19:02.080 Now this also reminds me this is not I mean, 0:19:02.200 --> 0:19:05.560 I guess it memory service. Maybe it did speak. But 0:19:06.520 --> 0:19:09.400 there was of course the Man Face Serpent God glyicon 0:19:09.880 --> 0:19:13.000 of the second century that is often held up as 0:19:13.040 --> 0:19:16.680 being a hoax, like it was actually a puppet according 0:19:17.280 --> 0:19:21.920 to commentators. But I've always wondered what to make of that, 0:19:22.040 --> 0:19:26.119 because it kind of if someone is performing puppetry and 0:19:26.160 --> 0:19:29.560 people are having an emotional or even religious reaction to it, 0:19:29.560 --> 0:19:31.840 it kind of depends how it's presented. Right, are you 0:19:31.920 --> 0:19:35.800 presenting Glycan the man Face Serpent as like this is it? 0:19:35.960 --> 0:19:39.080 This is an actual man face serpent God. Come take 0:19:39.080 --> 0:19:42.320 a look that its life? Is proof that he is real? 0:19:43.040 --> 0:19:46.240 Or is it something else? Is it more like performance 0:19:46.400 --> 0:19:48.919 or is it more like reinterpretation? You know, because you 0:19:48.960 --> 0:19:53.480 have plenty of examples where people will carry out performances 0:19:53.480 --> 0:19:56.639 in which people dress as divine and semi divine figures. 0:19:57.160 --> 0:19:59.760 It's not supposed to be like, look at the proof here, 0:20:00.119 --> 0:20:03.200 here is this hero on the stage. This means God 0:20:03.240 --> 0:20:03.560 is real? 0:20:03.880 --> 0:20:06.560 Funny enough, I think Glicon was also written about by 0:20:06.640 --> 0:20:11.159 Lucian of Somesada. But the I guess the crucial question 0:20:11.320 --> 0:20:13.919 is like is there an attempt at trickery or not? 0:20:14.160 --> 0:20:17.440 Like do do you want the audience to believe there 0:20:17.520 --> 0:20:19.560 is not somebody behind the mask? 0:20:20.080 --> 0:20:23.159 Right? And you know that's interesting because that still kind 0:20:23.200 --> 0:20:25.679 of applies to a lot of what's going on in 0:20:25.680 --> 0:20:29.200 the world today with things like like chatbots and so forth. 0:20:29.240 --> 0:20:33.280 And you know this idea that if we you know, 0:20:34.000 --> 0:20:36.439 what is coming out of the box, what is coming 0:20:36.440 --> 0:20:40.360 out of the artificial head, and you know, we how 0:20:40.359 --> 0:20:43.199 are we interpreting it? And are we thinking there is 0:20:43.240 --> 0:20:46.639 something there that is not? So it's like on what 0:20:46.760 --> 0:20:49.800 level is there trickery? And then there is like interpretation 0:20:49.880 --> 0:20:53.480 of the trickery and so forth. But at any rate, 0:20:53.640 --> 0:20:55.520 I think, you know, some of these examples they proved 0:20:55.520 --> 0:20:58.080 that well before people could make any kind of a 0:20:58.119 --> 0:21:01.640 mechanical thing, be it ahead or not ahead, that could speak, 0:21:01.920 --> 0:21:04.480 we were still capable of dreaming about it. And I 0:21:04.480 --> 0:21:07.280 think there's ample evidence that long before anyone attempted to 0:21:07.320 --> 0:21:10.680 make a head that could talk through mechanical means, individuals 0:21:10.680 --> 0:21:14.720 sought and sometimes found a voice emerging from disembodied heads, 0:21:15.000 --> 0:21:18.600 either real ones, you know, the remains of human beings 0:21:19.080 --> 0:21:24.200 or other animals, or likenesses of human heads, either attached 0:21:24.320 --> 0:21:27.640 or detached from statues, and so forth. And I think 0:21:27.640 --> 0:21:30.159 there's room between trickery and belief, you know, for the 0:21:30.200 --> 0:21:35.200 suspension of belief and ritual as well to take into account. 0:21:42.400 --> 0:21:45.160 But of course later on people would end up building real, 0:21:45.320 --> 0:21:49.560 operable machines that were at least attempting to produce speech 0:21:49.680 --> 0:21:51.480 that could be understood by humans. 0:21:51.960 --> 0:21:54.160 That's right, And this is where we get more into 0:21:54.200 --> 0:21:58.600 the deconstruction of what human speech is, which in and 0:21:58.640 --> 0:22:03.960 of itself as a whole subject, but there are key 0:22:04.040 --> 0:22:06.960 moments where we see some major advancements being made here. 0:22:07.560 --> 0:22:10.680 So another major entry to discuss in all of this 0:22:10.760 --> 0:22:13.520 is the work of German born Russian doctor, physicist and 0:22:13.560 --> 0:22:18.360 engineer Christian Gottlieb Kratzenstein, who lives seventeen twenty three through 0:22:18.400 --> 0:22:22.840 seventeen ninety five. So he was a man of various interests, 0:22:22.840 --> 0:22:26.200 including the use of electricity and medicine, and at the 0:22:26.240 --> 0:22:29.159 Saint Petersburg Science Academy at one point offered a prize 0:22:29.240 --> 0:22:33.879 for advancements made in researching the mechanisms behind the vowels 0:22:33.960 --> 0:22:37.800 AEI own you in human speech. So in seventeen seventy 0:22:37.880 --> 0:22:42.600 nine he presented his vowel organ to the university. The 0:22:43.640 --> 0:22:47.480 vowel organ consisted of a series of resonators that produced 0:22:47.600 --> 0:22:52.520 vowel like sounds on a constant pitch when excited by 0:22:52.560 --> 0:22:57.840 a read. I found some illustrations of these basic resonators 0:22:58.440 --> 0:23:05.320 via the UCL Psychology and Language Sciences Department. Here I 0:23:05.359 --> 0:23:11.440 also found a website linked at this website where you 0:23:11.480 --> 0:23:15.080 can find instructions for how to make your own resonators 0:23:15.080 --> 0:23:19.159 out of plumbing supplies, which I found rather insightful. I 0:23:19.200 --> 0:23:22.639 did not attempt it, but if you're into plumbing supplies 0:23:23.000 --> 0:23:26.560 and vowel sounds, it seems like a natural craft choice. 0:23:26.840 --> 0:23:30.000 But the key insight being here that by changing the 0:23:30.119 --> 0:23:33.520 shape of a physical resonating cavity, you can change the 0:23:33.680 --> 0:23:35.840 sound of the vowel produced. 0:23:36.280 --> 0:23:39.560 Right right. Another take on this, I was reading the 0:23:39.560 --> 0:23:42.720 BBC Future article The Machines That Learned to Listen by 0:23:42.800 --> 0:23:47.480 Kadia Muskvich, and it describes these as resonance tubes connected 0:23:47.520 --> 0:23:50.919 to organ pipes. So you know, this is not to 0:23:50.960 --> 0:23:52.520 say that we had. This is not on this like 0:23:52.560 --> 0:23:55.120 the same level as some sort of imaginary brazen head 0:23:55.160 --> 0:23:57.280 that's going to speak of its own and spout out, 0:23:57.520 --> 0:23:59.879 spit out wisdom for you to interpret. This is a 0:24:00.320 --> 0:24:04.679 just figuring out how these vowel sounds are produced and 0:24:04.720 --> 0:24:10.600 reproducing them through a basic mechanical system. Mosfitch also points 0:24:10.600 --> 0:24:12.760 out a few other key individuals in the advancement of 0:24:12.800 --> 0:24:17.560 this technology. There's Wolfgang von Kimplin in Vienna, who created 0:24:17.600 --> 0:24:22.439 a similar acoustic mechanical speech machine about ten years after Kratzenstein. 0:24:23.160 --> 0:24:27.280 And then she also mentions English inventor Charles Wheatstone, who 0:24:27.320 --> 0:24:29.600 would improve on this in the early nineteenth century. 0:24:30.000 --> 0:24:32.720 Charles Wheatstone. I'm going to mention him again in a minute, 0:24:32.760 --> 0:24:35.560 but he's also notable because he was one of the 0:24:35.640 --> 0:24:41.000 inventors of the first commercially successful form of the telegraph. 0:24:41.480 --> 0:24:43.600 So we talked about him in our episode on You 0:24:44.480 --> 0:24:47.200 of the Telegraph. But when it comes to the one 0:24:47.480 --> 0:24:50.240 you mentioned before, that von Kempelan's machine, this is interesting 0:24:50.240 --> 0:24:57.119 because I read that while this machine was allegedly real, 0:24:57.200 --> 0:24:59.639 it was a real attempt to make a machine that 0:24:59.680 --> 0:25:04.920 would Von Kimpelan is now known for essentially being a hoaxer, 0:25:05.000 --> 0:25:08.879 because he tried to create other automata, including a chess 0:25:08.960 --> 0:25:12.119 playing automaton that was actually a hoax. It had a 0:25:12.200 --> 0:25:14.879 human inside it doing the move, so it was a 0:25:14.920 --> 0:25:16.159 fake robot. 0:25:16.240 --> 0:25:19.480 Though as a fake still really impressive. It's interesting where 0:25:19.520 --> 0:25:21.600 you get in, like what sometimes you're wondering you have 0:25:21.640 --> 0:25:24.600 to wonder what the line is between, you know, the 0:25:24.640 --> 0:25:28.560 actual technological innovation and trickery. I mean, obviously it's deception, 0:25:28.960 --> 0:25:31.720 and if you have a secret chamber in which there's 0:25:31.760 --> 0:25:34.360 a whole person doing stuff, you know, that's a real 0:25:34.400 --> 0:25:37.680 red flag there as well. But still the trickery is 0:25:37.720 --> 0:25:38.720 pretty ingenious too. 0:25:39.200 --> 0:25:42.439 Yeah. Well yeah, I mean it takes skill to be 0:25:42.440 --> 0:25:43.160 a good magician. 0:25:43.520 --> 0:25:44.000 Yeah. 0:25:44.080 --> 0:25:46.760 Anyway, this brings us to the example that I was 0:25:46.760 --> 0:25:50.000 really excited to talk about in today's episode, which is 0:25:50.040 --> 0:25:55.400 the speaking machine of a nineteenth century inventor named Joseph 0:25:55.680 --> 0:25:59.199 fober So. One of my main sources here is just 0:25:59.320 --> 0:26:02.440 generally a good source on the history of speech synthesis 0:26:02.520 --> 0:26:06.560 and talking machines. It was a book chapter in the 0:26:06.640 --> 0:26:10.680 Rutledge Handbook of Phonetics from twenty nineteen by an author 0:26:10.760 --> 0:26:14.560 named Brad H. Story, who is part of the faculty 0:26:14.640 --> 0:26:17.320 of the Department of Speech, Language and Hearing Sciences at 0:26:17.359 --> 0:26:22.560 the University of Arizona, and Story in this chapter traces 0:26:22.640 --> 0:26:26.159 the history of speech synthesis from the mechanical methods of 0:26:26.200 --> 0:26:29.680 the eighteenth and nineteenth centuries to the digital techniques of 0:26:29.720 --> 0:26:31.920 the present. So it's the whole sort of modern arc 0:26:32.119 --> 0:26:34.920 of these machines. But the thing I really want to 0:26:34.960 --> 0:26:37.560 focus in on here now is this machine that I 0:26:37.600 --> 0:26:41.919 mentioned a minute ago by the nineteenth century German inventor 0:26:42.119 --> 0:26:45.800 Joseph Fober. This features heavily at the beginning of stories 0:26:45.880 --> 0:26:49.520 chapter here. So this machine was at various different times 0:26:49.680 --> 0:26:53.760 called the marvelous talking machine. You got a hyphen between 0:26:53.800 --> 0:26:58.399 talking machine and also the euphonia from the Greek meaning 0:26:58.480 --> 0:27:02.639 good sound or sweet sound. We'll see about that as 0:27:02.680 --> 0:27:06.480 we as we go on. Robi included one illustration of 0:27:06.520 --> 0:27:08.320 the machine for you to look at here. I think 0:27:08.359 --> 0:27:10.960 this may have been from some kind of promotional material 0:27:11.040 --> 0:27:13.440 when this machine was featured in an exhibit that I'll 0:27:13.480 --> 0:27:14.280 describe in a bit. 0:27:14.640 --> 0:27:18.040 I love it in part because right there is this 0:27:18.160 --> 0:27:24.000 angelic human face like right there on the machine, seemingly 0:27:24.160 --> 0:27:28.000 as like decoration or maybe tribute. I'm not sure, but 0:27:28.400 --> 0:27:31.400 I'm not sure if it's actually necessary to the mechanics 0:27:31.400 --> 0:27:31.919 of the device. 0:27:31.920 --> 0:27:35.320 Here, I think it sort of is. Well, I'll explain. 0:27:35.840 --> 0:27:40.960 So story introduces Fober's machine through the eyes of another 0:27:41.200 --> 0:27:44.879 inventor and scientist of the day named Joseph Henry. A 0:27:44.880 --> 0:27:50.160 different Joseph, a researcher on electromagnetic induction and also the 0:27:50.200 --> 0:27:56.560 inaugural secretary of the Smithsonian Institution. Henry encountered Faber's marvelous 0:27:56.600 --> 0:28:01.520 talking machine at a private exhibition in Philadelphia on December twentieth, 0:28:01.640 --> 0:28:06.159 eighteen forty five, and he described the demonstration in a 0:28:06.280 --> 0:28:09.480 letter to a colleague named HM. Alexander. So we have 0:28:09.680 --> 0:28:12.800 contemporaneous notes on what it was doing and what it 0:28:12.840 --> 0:28:16.639 looked like in this private demonstration. So here's how it worked. 0:28:17.320 --> 0:28:22.359 It was controlled by an operator via a mainly by 0:28:22.520 --> 0:28:26.480 foot pedals and a keyboard, essentially just like an organ, 0:28:26.640 --> 0:28:29.680 like a chamber organ, and in fact the device could 0:28:29.720 --> 0:28:33.359 in some ways be considered a modified organ. So you 0:28:33.400 --> 0:28:36.679 had a foot pedal that operated a bellows and that 0:28:36.720 --> 0:28:41.040 would supply airflow to the whole system, and the bellows 0:28:41.120 --> 0:28:45.120 pumped air through an artificial larynx that had vocal cords 0:28:45.200 --> 0:28:47.000 that were in this source said to be made of 0:28:47.120 --> 0:28:53.080 rubber and these so this artificial glottis or artificial vocal 0:28:53.120 --> 0:28:57.719 cords would vibrate to produce the fundamental sound of the 0:28:57.760 --> 0:29:00.840 machine's voice when air was flowing through them. And then 0:29:00.880 --> 0:29:05.120 you had sixteen keys on the keyboard which were connected 0:29:05.120 --> 0:29:09.720 by strings and levers to the various components that controlled 0:29:09.920 --> 0:29:12.719 the shaping of that sound of that, you know, the 0:29:12.720 --> 0:29:16.560 resonating sound from that airflow through the glottis into speech. 0:29:17.040 --> 0:29:19.600 One of the interesting things is, as we've been saying, 0:29:19.640 --> 0:29:22.560 this device actually had a face. So the face was 0:29:22.640 --> 0:29:26.520 made of carved wood, essentially a large doll head, but 0:29:26.600 --> 0:29:29.400 it had a hinged jaw. So maybe you should think 0:29:29.400 --> 0:29:32.160 of it more like a ventriloquist dummy. You're loving this, 0:29:32.320 --> 0:29:35.880 aren't you. Yeah, Night of the Living dummy. But it 0:29:36.160 --> 0:29:40.560 can actually speak. And so inside the dummy's mouth there 0:29:40.640 --> 0:29:44.959 was an ivory tongue that could be moved around inside 0:29:44.960 --> 0:29:49.200 the oral cavity to control the shape of the resonating chamber. 0:29:49.920 --> 0:29:52.920 And by controlling these different elements like the mouth and 0:29:52.960 --> 0:29:56.000 the tongue and all that, with the keys on the keyboard. 0:29:57.160 --> 0:30:01.280 It quote imposed time varying change to the air cavity 0:30:01.640 --> 0:30:08.320 appropriate for generating apparently convincing renditions of connected speech. So 0:30:08.400 --> 0:30:11.520 it may not have sounded perfect or even pleasant, but 0:30:11.800 --> 0:30:15.640 apparently people in the room could understand what the machine 0:30:15.680 --> 0:30:19.040 was saying when Fober operated it. So this is eighteen 0:30:19.120 --> 0:30:23.360 forty five and the machine is speaking intelligible words. Henry 0:30:23.840 --> 0:30:27.200 in this letter compares it favorably to a different talking machine, 0:30:27.200 --> 0:30:29.360 one he had seen years before. This was one of 0:30:29.400 --> 0:30:31.640 the ones you mentioned, Rob, the one built by the 0:30:31.680 --> 0:30:35.960 English scientist and inventor Charles Wheatstone, again the telegraph guy. 0:30:36.880 --> 0:30:41.920 Wheatstone's talking machine was capable of being understood for the 0:30:42.040 --> 0:30:45.360 set of words it could produce, but Fober's machine was 0:30:45.600 --> 0:30:50.120 far superior because its speech repertoire was infinitely variable, so 0:30:50.160 --> 0:30:54.280 he could speak whole sentences, and those sentences could contain 0:30:54.400 --> 0:30:57.120 any words and any sounds he wanted, as long as 0:30:57.120 --> 0:31:00.080 they were in one of the covered languages. Obviously it 0:31:00.120 --> 0:31:03.880 couldn't do, you know, like tonal languages, or like speak 0:31:03.920 --> 0:31:06.760 Mandarin or something, But it seems like mainly it was 0:31:06.760 --> 0:31:09.280 speaking German and English. It was said at the time 0:31:09.320 --> 0:31:12.200 that it could speak any European language. Now, I think 0:31:12.240 --> 0:31:14.560 one thing that's really worth noting here is that if 0:31:14.640 --> 0:31:17.440 you imagine how a machine like this would work, the 0:31:17.680 --> 0:31:23.240 success of the performance would depend heavily on the skill 0:31:23.600 --> 0:31:27.000 of the operator, since the speech patterns are not like 0:31:27.200 --> 0:31:32.360 programmed and you know, not sort of expressed automatically, but 0:31:32.560 --> 0:31:36.880 expressed in real time by the player operating the bellows 0:31:36.920 --> 0:31:38.760 and the keys. And I think also there were some 0:31:39.240 --> 0:31:42.560 screws and stuff that would manipulate pitch and things like that, 0:31:42.680 --> 0:31:45.280 So you have to play this just like you would 0:31:45.280 --> 0:31:49.560 play a musical instrument. So different players using the same 0:31:49.640 --> 0:31:54.200 machine would probably produce fairly different sounding speech, even if 0:31:54.200 --> 0:31:57.760 they had memorized which keys corresponded to which phonetic units. 0:31:58.280 --> 0:32:01.480 So nobody ever read says this, But you know, I'm 0:32:01.560 --> 0:32:04.800 kind of picturing Fober as a sort of phantom of 0:32:04.800 --> 0:32:07.040 the opera at the at the organ keyboard. You know, 0:32:07.080 --> 0:32:09.040 he's not just like pressing the keys, but giving a 0:32:09.040 --> 0:32:12.880 real passionate and dramatic performance. When somebody sells it, he 0:32:12.960 --> 0:32:15.640 yells at yeah, make it say como tale vou or whatever. 0:32:16.840 --> 0:32:19.120 It also sang songs, by the way, I'll get into 0:32:19.120 --> 0:32:21.600 that in a minute. But I was wondering, what did 0:32:21.640 --> 0:32:24.120 what are people asking, you know, what's the equivalent in 0:32:24.160 --> 0:32:27.160 eighteen forty five of yelling out, you know, play Freebird? 0:32:27.200 --> 0:32:29.040 And I was thinking, maybe it's people are yelling for 0:32:29.120 --> 0:32:30.400 TYPICANU and Tyler too. 0:32:31.000 --> 0:32:32.520 Oh yeah. 0:32:32.600 --> 0:32:35.520 So an interesting detail that story includes in this chapter 0:32:35.720 --> 0:32:38.760 is that this was not the first time Fober had 0:32:38.800 --> 0:32:41.640 built a talking machine. In fact, this was not the 0:32:41.680 --> 0:32:45.680 first time Fober had built this exact talking machine. There 0:32:45.720 --> 0:32:48.600 was an earlier version of it that was destroyed by 0:32:48.760 --> 0:32:52.640 Fober himself, quote in a bout of depression and intoxication. 0:32:53.520 --> 0:32:56.000 I should say that nearly every source I read on 0:32:56.080 --> 0:33:01.920 Fober mentions something about him being disheveled or haunted, obsessed 0:33:02.000 --> 0:33:05.360 with his machine, and generally emotionally unwell or at the 0:33:05.440 --> 0:33:07.880 very least having a really rough time a lot of 0:33:07.920 --> 0:33:11.920 the time. Multiple writers describe him in terms containing a 0:33:11.920 --> 0:33:17.160 lot of pity. But so, it took Fober apparently twenty 0:33:17.360 --> 0:33:20.560 years to perfect the first version of the machine, the 0:33:20.600 --> 0:33:24.160 one that he drunkenly destroyed, but he was able to 0:33:24.400 --> 0:33:27.680 recreate the second version within a year of that. And 0:33:27.720 --> 0:33:30.640 this kind of suggests to me the possibility that the 0:33:30.680 --> 0:33:34.560 original creation of the machine may have really been a 0:33:34.560 --> 0:33:38.360 project of fundamental research about phonetics more than it was 0:33:38.400 --> 0:33:41.920 about engineering. And so once he had the knowledge in 0:33:42.040 --> 0:33:44.800 hand of how each sound was produced, like what the 0:33:44.880 --> 0:33:48.360 shape of the oral cavity, you know, how that corresponded 0:33:48.360 --> 0:33:51.880 to the sounds, recreating the machine itself might have been 0:33:51.920 --> 0:33:54.800 a relatively simple proposition. Is really what you needed was 0:33:54.840 --> 0:33:58.800 the knowledge about how phonetics correspond to physical shapes. 0:33:59.360 --> 0:34:02.239 Yeah, had that, and certainly they had notes on the 0:34:02.240 --> 0:34:06.120 matter and his designs recorded. It would be easier to 0:34:06.160 --> 0:34:07.400 come back and reproduce that. 0:34:07.840 --> 0:34:13.520 Yeah. So Joseph Henry's letter about Faber's talking machine demonstration 0:34:13.840 --> 0:34:17.440 it also includes speculation about the uses to which a 0:34:17.480 --> 0:34:21.000 machine like this could be put. One interesting idea he 0:34:21.080 --> 0:34:24.240 has is what if you could take a spoken message 0:34:24.280 --> 0:34:29.400 at one location and code that spoken message into inputs 0:34:29.400 --> 0:34:34.600 on this keyboard on this machine, and then, through electromagnetic means, 0:34:35.080 --> 0:34:40.520 transmit those keystrokes across wires to a totally separate second location, 0:34:41.239 --> 0:34:45.040 and then those electrical signals could operate the speech organs 0:34:45.080 --> 0:34:48.600 of the doll faced machine. In the second location. You 0:34:48.640 --> 0:34:55.080 would essentially be transmitting speech itself. Across great distance. Notable 0:34:55.080 --> 0:34:58.640 that Henry's idea here is roughly thirty years before Alexander 0:34:58.680 --> 0:35:02.120 Graham Bell demonstrates the principle of the telephone. But there 0:35:02.160 --> 0:35:06.000 is a very important difference, which is that while Bell's 0:35:06.000 --> 0:35:09.480 telephone and these are stories words here quote transmitted an 0:35:09.560 --> 0:35:14.680 electrical analog of the speech pressure wave. Henry's description alluded 0:35:14.719 --> 0:35:20.000 to representing speech in compressed form based on slowly varying 0:35:20.080 --> 0:35:23.879 movements of the operator's hands, fingers, and feet as they 0:35:23.920 --> 0:35:27.920 formed the keystroke sequences required to produce an utterance, a 0:35:27.960 --> 0:35:31.719 signal processing technique that would not be implemented into telephone 0:35:31.719 --> 0:35:36.160 transmission systems for nearly another century. So the interesting thing 0:35:36.200 --> 0:35:39.839 about Henry here is that he's not just imagining converting 0:35:39.920 --> 0:35:43.200 the sound of a voice into an impulse that travels 0:35:43.239 --> 0:35:47.319 along the wire. He's imagining a coding process. It's put 0:35:47.400 --> 0:35:50.719 into code for the transmission and then decoded by the 0:35:50.760 --> 0:35:51.960 machine at the other end. 0:35:52.640 --> 0:35:56.160 I can't help but try to imagine this alternate past 0:35:56.440 --> 0:36:00.560 in which instead of early telephones, people all had this 0:36:01.320 --> 0:36:06.120 weird cherub head mounted on the wall that then speaks 0:36:06.160 --> 0:36:09.920 to you. In this I'm assumed slightly haunting voice. 0:36:10.160 --> 0:36:12.160 Oh, I'll get to the haunting voice in a second, 0:36:12.960 --> 0:36:16.319 but anyway, story flags It as historically significant that this 0:36:16.400 --> 0:36:21.040 one invention had both succeeded in producing generally intelligible synthetic 0:36:21.080 --> 0:36:23.880 speech to people in the room with it, and it 0:36:23.960 --> 0:36:27.800 had inspired at least one onlooker to start considering ideas 0:36:27.800 --> 0:36:31.200 for the electrical transmission of low bandwidth speech from one 0:36:31.239 --> 0:36:36.160 place to another. But neither of these possibilities really went anywhere. 0:36:36.360 --> 0:36:39.239 Henry did not devote any more effort to musing about 0:36:39.280 --> 0:36:44.440 the electrical transmission, and Fober's machine ended up being a 0:36:44.440 --> 0:36:49.760 circus side show, almost literally. So after this, Fober needed money, 0:36:50.280 --> 0:36:53.840 and beginning in eighteen forty six, to get money, he 0:36:54.000 --> 0:36:57.360 signed on to demonstrate his machine for p. T. Barnum 0:36:57.480 --> 0:37:00.959 got to have something for everybody, even people who want 0:37:01.000 --> 0:37:05.080 a talking doll head operated by a disheveled German organ master. 0:37:06.280 --> 0:37:10.239 So Faber committed to exhibit the marvelous Speaking Machine four 0:37:10.320 --> 0:37:14.239 Barnum at the Egyptian Hall in London. This was like 0:37:14.360 --> 0:37:18.200 a general exhibition hall in Piccadilly, which hosted all kinds 0:37:18.239 --> 0:37:20.799 of shows, but I think especially in the latter part 0:37:20.840 --> 0:37:23.920 of the nineteenth century, it was known for showing like 0:37:23.960 --> 0:37:29.479 a lot of Mountebanks and fraudulent spiritualist demonstrators. Yeah, I'll 0:37:29.520 --> 0:37:32.920 reveal to you that you're actually a reincarnation of Cleopatra. 0:37:34.320 --> 0:37:37.600 Lucky you. But by noting that that's just a random thing, 0:37:37.600 --> 0:37:40.719 I'm not trying to cast dispersions on Fober, because I 0:37:40.719 --> 0:37:43.680 want to stress that it seems totally clear that Fober 0:37:43.960 --> 0:37:46.640 was no con artist. As best we can tell, his 0:37:46.760 --> 0:37:51.280 machine really did work, and when played correctly, it did 0:37:51.400 --> 0:37:54.879 really speak original sentences that people could, for the most 0:37:54.880 --> 0:37:56.000 part understand. 0:37:56.360 --> 0:37:56.640 Though. 0:37:56.800 --> 0:38:00.799 One thing that emerges from reading descriptions of this is 0:38:00.840 --> 0:38:07.120 that coding intelligible information and sounding like speech are two 0:38:07.160 --> 0:38:10.640 completely different things. So it seems that a lot of 0:38:10.680 --> 0:38:14.480 people could tell what the machine was saying, but still 0:38:14.560 --> 0:38:18.480 they were not very impressed by what they heard. And 0:38:18.560 --> 0:38:23.400 I found a spectacularly evocative description of what the machine 0:38:23.440 --> 0:38:26.560 was like I recorded in a book called Instruments and 0:38:26.600 --> 0:38:30.600 the Imagination by Thomas L. Hankins and Robert J. Silverman, 0:38:30.760 --> 0:38:34.319 Princeton University Press, nineteen ninety nine. But the main thing 0:38:34.360 --> 0:38:37.319 here is that they're quoting a person who saw the 0:38:37.360 --> 0:38:40.200 machine in person in eighteen forty six, I believe, and 0:38:40.440 --> 0:38:42.960 then wrote about it in a memoir. But generally the 0:38:43.000 --> 0:38:45.520 authors here they note that there were like some satirical 0:38:45.640 --> 0:38:49.640 articles making reference to Faber's machine, suggesting, for example, that 0:38:49.760 --> 0:38:52.319 it could be used to replace the speaker of the 0:38:52.400 --> 0:38:57.319 House of Commons. Yak, yeah, those wacky politicians. But then 0:38:57.960 --> 0:38:59.720 well they do kind of make a funny point. Actually, 0:38:59.719 --> 0:39:01.839 they say, like you could just program it to say 0:39:02.040 --> 0:39:04.080 order order at ten minute intervals. 0:39:06.040 --> 0:39:07.760 Well that's pretty good, that's funny today. 0:39:08.200 --> 0:39:10.960 Yeah. But anyway, then there's a part of the book 0:39:11.000 --> 0:39:15.000 where they're including this evocative written account which is from 0:39:15.080 --> 0:39:20.120 a London theater manager named John Hollingshead who saw this 0:39:20.200 --> 0:39:22.960 machine in person when he was nineteen years old and 0:39:23.000 --> 0:39:25.880 then wrote about it in a memoirs or some book. 0:39:26.120 --> 0:39:30.760 But anyway, this is Hollingshead's account. The exhibitor, Professor Fober, 0:39:31.239 --> 0:39:34.880 was a sad faced man, dressed in respectable, well worn 0:39:34.920 --> 0:39:39.000 clothes that were soiled by contact with tools, wood and machinery. 0:39:39.600 --> 0:39:43.160 The room looked like a laboratory and workshop, which it was. 0:39:43.800 --> 0:39:46.560 The professor was not too clean, and his hair and 0:39:46.600 --> 0:39:50.239 beard sadly wanted the attention of a barber. I have 0:39:50.320 --> 0:39:53.000 no doubt that he slept in the same room as 0:39:53.040 --> 0:39:57.480 his figure, his scientific Frankenstein Monster. Note. I guess the 0:39:57.520 --> 0:40:00.560 novel would have only been a few decades old this time. 0:40:00.960 --> 0:40:03.359 Yeah, yeah, eighteen eighteen on Frankenstein there. 0:40:03.600 --> 0:40:07.200 Yeah, sorry, going on with hallings Head and I felt 0:40:07.239 --> 0:40:10.000 the secret influence of an idea that the two were 0:40:10.080 --> 0:40:12.160 destined to live and die together. 0:40:12.760 --> 0:40:15.800 Oh my god, this is those pretty strong words. 0:40:16.160 --> 0:40:20.319 Yes, the professor, with a slight German accent, put his 0:40:20.440 --> 0:40:24.440 wonderful toy in motion. He explained its action. It was 0:40:24.480 --> 0:40:28.520 not necessary to prove the absence of deception. One keyboard 0:40:28.640 --> 0:40:32.680 touched by the professor produced words which, slowly and deliberately, 0:40:32.800 --> 0:40:36.480 in a hoarse, sepulchral voice, came from the mouth of 0:40:36.520 --> 0:40:39.200 the figure, as if from the depths of a tomb. 0:40:39.760 --> 0:40:43.040 It wanted little imagination to make the very few visitors 0:40:43.120 --> 0:40:46.959 believe that the figure contained an imprisoned human or half 0:40:47.080 --> 0:40:51.560 human being, bound to speak slowly when tormented by the 0:40:51.640 --> 0:40:55.680 unseen power outside. No one thought for a moment that 0:40:55.719 --> 0:40:58.160 they were being fooled by a second edition of The 0:40:58.360 --> 0:41:02.200 Invisible Girl fra Allod And by the way, the reference 0:41:02.239 --> 0:41:04.719 to the Invisible Girl fraud, I believe is about the 0:41:05.080 --> 0:41:08.680 many fake machines and fake automata that were actually worked 0:41:08.719 --> 0:41:11.919 by having a human hidden inside operating it but going 0:41:11.960 --> 0:41:14.880 out so holling said, says, nobody thought that there was 0:41:14.920 --> 0:41:17.920 an invisible girl operating. This as clear, this is real. 0:41:18.200 --> 0:41:22.120 He goes on, there were truth, laborious invention, and good 0:41:22.200 --> 0:41:25.719 faith in every part of the melancholy room. As a 0:41:25.760 --> 0:41:29.960 crowning display, the head sang a sepulchral version of God 0:41:30.080 --> 0:41:34.640 Save the Queen, which suggested, inevitably, God save the inventor. 0:41:35.280 --> 0:41:39.279 This extraordinary effect was achieved by the professor working two keyboards, 0:41:39.600 --> 0:41:43.040 one for the words and one for the music. Never 0:41:43.160 --> 0:41:46.880 probably before or since, has the national anthem been so sung, 0:41:47.520 --> 0:41:51.360 sadder and wiser. I and the few visitors crept slowly 0:41:51.440 --> 0:41:54.440 from the place, leaving the professor with his one and 0:41:54.520 --> 0:41:58.920 only treasure, his child of infinite labor and unmeasurable sorrow. 0:41:59.320 --> 0:42:03.960 Oh wow, that is a lot. I mean, obviously, he's 0:42:04.040 --> 0:42:06.560 lee lays it on really thick about the sadness of 0:42:06.600 --> 0:42:09.600 the inventor here. And then also there's the ideas like 0:42:09.640 --> 0:42:12.520 this was no hoax, this was real and it was depressing. 0:42:12.960 --> 0:42:16.719 Yeah, I thought, it's a weird mix of like like 0:42:16.920 --> 0:42:20.920 pity but real admiration, you know that, Like, there's something 0:42:21.280 --> 0:42:24.640 beautiful and honest and true about this machine and his 0:42:24.680 --> 0:42:27.799 devotion to it and the genius it took to create it. 0:42:28.000 --> 0:42:32.040 But also it makes everybody feel bad and nobody wants 0:42:32.080 --> 0:42:34.040 to look at it or listen to it, and everybody 0:42:34.120 --> 0:42:38.560 leaves feeling depressed. Yeah, something about that struck me as 0:42:38.560 --> 0:42:41.879 actually quite poignant and meaningful. Maybe we can come back 0:42:41.880 --> 0:42:43.840 to that in a minute, but I did want to 0:42:43.840 --> 0:42:47.520 flag that there was one notable visitor who, coming back 0:42:47.520 --> 0:42:51.640 to the Invisible Girl suspicion, he did at first suspect fraud, 0:42:51.760 --> 0:42:54.960 and that was the Duke of Wellington. I was reading 0:42:54.960 --> 0:42:57.240 about this in a book called The Shows of London 0:42:57.320 --> 0:43:02.000 by Richard Daniel Atlick, and at Lick recounts that Wellington, 0:43:02.320 --> 0:43:05.319 when he first went to the demonstration, he was so 0:43:05.480 --> 0:43:09.719 impressed by Faber's speaking machine that he asked to be 0:43:09.840 --> 0:43:12.600 allowed to touch the keys with his own fingers, you know, 0:43:12.640 --> 0:43:15.359 so he could see that it was genuine. And then 0:43:15.520 --> 0:43:17.840 he did confirm that it was genuine, and then he 0:43:17.920 --> 0:43:20.799 insisted that he'd be taught how to use it, so 0:43:21.040 --> 0:43:24.080 Fober taught the Duke to play the machine in both 0:43:24.160 --> 0:43:28.120 German and English, and Wellington did get it like he could. 0:43:28.160 --> 0:43:30.640 He could make it speak sentences in German and English, 0:43:30.800 --> 0:43:33.680 and he was amazed, writing in the visitor's log of 0:43:33.760 --> 0:43:37.840 the exhibit that the speaking machine, or the Euphonia, was 0:43:37.920 --> 0:43:52.600 quote an extraordinary production of mechanical genius. Faber's machine also 0:43:52.640 --> 0:43:56.400 got rave reviews in The Times, in the Illustrated London News. 0:43:56.440 --> 0:44:00.160 A lot of people like looked at it and they 0:44:00.000 --> 0:44:03.000 they thought that, like, yeah, this is a work of genius. 0:44:03.239 --> 0:44:06.560 It's incredible that he's done this. But at the same time, 0:44:07.120 --> 0:44:11.279 audiences really were not into it. Barnum himself noticed that 0:44:11.320 --> 0:44:14.600 Fober's machine was not attracting crowds, it was not selling 0:44:14.640 --> 0:44:19.040 tickets and not generating revenue, and so eventually he took 0:44:19.160 --> 0:44:23.560 Fober's machine out of the Egyptian Hall in London and 0:44:23.760 --> 0:44:26.360 added it to a traveling exhibit that went around the 0:44:26.360 --> 0:44:31.440 English countryside doing performances. And from here Fober himself seems 0:44:31.440 --> 0:44:34.880 to kind of disappear from the historical record. Some sources 0:44:34.920 --> 0:44:38.640 indicate that he may have died by suicide during this period, 0:44:38.880 --> 0:44:42.720 though that isn't known. For sure, But after historical sources 0:44:42.760 --> 0:44:46.440 stopped mentioning Fober himself, they still make references to his 0:44:46.480 --> 0:44:51.240 machine reading from story here quote. Although his talking machine 0:44:51.520 --> 0:44:54.560 continued to make side show like appearances in Europe and 0:44:54.640 --> 0:44:58.600 North America over the next thirty years. It seems a relative, 0:44:58.719 --> 0:45:01.560 perhaps a niece or nephew you may have inherited the 0:45:01.600 --> 0:45:04.160 machine and performed with it to generate income. 0:45:05.160 --> 0:45:08.400 So maybe no matter whatever happened to him, maybe a 0:45:08.440 --> 0:45:11.719 relative with a little more showmanship like stepped in and 0:45:11.920 --> 0:45:14.239 was able to make at least some sort of an 0:45:14.239 --> 0:45:15.040 income off of it. 0:45:15.560 --> 0:45:18.240 Yes, but then again, like I'm struck by the strange, 0:45:18.280 --> 0:45:23.600 ironic sadness of this. This this was actually a scientifically 0:45:23.680 --> 0:45:27.520 significant invention, Like he had done something kind of amazing, 0:45:28.640 --> 0:45:31.880 but it just never really went anywhere under his mastery. 0:45:31.920 --> 0:45:35.520 And then yeah, maybe a relative was a better Carnival 0:45:35.560 --> 0:45:39.520 Barker essentially to perform with the machine and make some 0:45:39.560 --> 0:45:40.319 money off of it. 0:45:40.520 --> 0:45:42.799 I mean, it reminds me of so many advancements in 0:45:43.000 --> 0:45:48.480 say robotics, that we've seen over the years, where oftentimes, 0:45:48.719 --> 0:45:51.480 you know, to a certain extent, unfairly, they'll just be 0:45:51.600 --> 0:45:54.160 one little clip of it that goes viral and people 0:45:54.200 --> 0:45:57.360 react to be it some sort of you know, a 0:45:58.120 --> 0:46:01.480 human likeness with facial features that seem to be moving 0:46:01.600 --> 0:46:04.520 or operating in an uncanny way, or something like the 0:46:04.600 --> 0:46:11.160 various dog robots from Boston Dynamics that are very impressive 0:46:11.200 --> 0:46:14.560 but also may be interpreted as being a bit creepy. 0:46:14.880 --> 0:46:17.000 And so even though they are these, you know, they're 0:46:17.000 --> 0:46:23.000 often examples of a real impressive technological advancement setting aside 0:46:23.320 --> 0:46:26.719 actual applications, you can have a situation where something like 0:46:26.760 --> 0:46:32.919 that is not as comforting, not as entertaining as say 0:46:32.920 --> 0:46:35.960 an act of puppetry, or even an act of just 0:46:36.320 --> 0:46:40.080 outright well maybe not fraud, but say a robot or 0:46:40.120 --> 0:46:46.000 a costume depicting a robot maybe ultimately maybe more reassuring, 0:46:46.120 --> 0:46:48.399 maybe more fun compared to the actual thing. 0:46:48.920 --> 0:46:51.279 Well yeah, which may which may just be fun or 0:46:51.280 --> 0:46:54.080 may in fact be fraud, depending on what exactly they're 0:46:54.120 --> 0:46:54.839 saying about it. 0:46:55.200 --> 0:46:56.799 Yeah, but this. 0:46:56.760 --> 0:46:58.480 Is a great point and it brings me to I 0:46:58.520 --> 0:47:01.160 just wanted to mention a few of the the general 0:47:01.239 --> 0:47:05.200 notes about the history of speech synthesis from the end 0:47:05.200 --> 0:47:09.719 of this book chapter by Brad's story Story Rights that 0:47:09.800 --> 0:47:14.120 you know, while there are technological use cases for speech synthesizers. 0:47:14.560 --> 0:47:16.759 You know, we've got a number of them operating in 0:47:16.760 --> 0:47:21.080 consumer technology today, and even before you had you know, 0:47:21.360 --> 0:47:24.680 personal digital assistants and stuff, there would be use cases 0:47:24.719 --> 0:47:28.920 for speech synthesizers for, for example, people who have a 0:47:28.920 --> 0:47:32.400 disability that makes it difficult or impossible for them to speak. 0:47:32.800 --> 0:47:35.759 Another one is that apparently this was actually used by 0:47:35.800 --> 0:47:38.000 the Allies in World War Two, there were some forms 0:47:38.000 --> 0:47:42.880 of speech synthesis that would allow sort of covert coded 0:47:43.080 --> 0:47:46.600 transmissions of something like a phone call, so you could 0:47:46.600 --> 0:47:49.680 have a phone call between like FDR and Winston Churchill. 0:47:50.000 --> 0:47:52.600 It's not really a phone call, it's like a transmitted 0:47:53.200 --> 0:47:56.759 synthesized bit of speech, and so it's very secure. But 0:47:56.840 --> 0:47:59.919 it doesn't sound like the person talking. It sounds maybe 0:48:00.080 --> 0:48:03.320 more like the euphonia, kind of robotic and unnatural and 0:48:03.960 --> 0:48:07.920 maybe making the president's giggle a bit a president Prime minister. 0:48:09.200 --> 0:48:12.240 But anyway, So what story says is that a large 0:48:12.320 --> 0:48:15.839 number of these systems have actually been primarily used as 0:48:16.480 --> 0:48:20.600 research tools, as scientific tools for understanding the nature of 0:48:20.680 --> 0:48:25.960 human speech. It's by trying to reproduce human speech and 0:48:26.160 --> 0:48:29.919 failing at it that we come closer to understanding how 0:48:30.040 --> 0:48:34.000 speech actually works in the human body. But the second 0:48:34.040 --> 0:48:36.560 general observation that I thought is interesting, and this seems 0:48:36.600 --> 0:48:41.000 to be very much reflected in the Fober's machine example. 0:48:41.520 --> 0:48:44.319 It is much easier to create a machine that can 0:48:44.360 --> 0:48:50.440 speak intelligibly than one that can speak naturally. So that 0:48:50.480 --> 0:48:53.600 indicates that when we talk, there's actually more than one 0:48:53.719 --> 0:48:58.240 thing going on. Yes, we are conveying mental information coded 0:48:58.280 --> 0:49:02.480 in words, and the substance of that coding is phonetic. 0:49:02.640 --> 0:49:04.880 It's a series of sounds. But of course, you know, 0:49:04.920 --> 0:49:08.279 the ironic thing to people who were used to thinking 0:49:08.280 --> 0:49:11.319 about words as text is that the phonetic core of 0:49:11.400 --> 0:49:14.680 language long predates writing, so like the written text of 0:49:14.719 --> 0:49:17.759 a word is a visual code for the sound of 0:49:17.800 --> 0:49:21.000 the word, which is the code for its meaning. But anyway, 0:49:21.239 --> 0:49:24.520 so machines for hundreds of years have been able to 0:49:24.680 --> 0:49:29.440 produce more or less intelligible phonetic code. They can speak words, 0:49:29.440 --> 0:49:32.360 and people can understand what the words are supposed to be, 0:49:33.440 --> 0:49:37.239 But it doesn't necessarily mean that people perceive these machines 0:49:37.320 --> 0:49:42.000 as speaking, because there's another important quality to speech that 0:49:42.120 --> 0:49:45.200 was not really captured by these early machines, and you 0:49:45.239 --> 0:49:48.160 could argue is still somewhat lacking in the best speech 0:49:48.640 --> 0:49:52.400 synthesis of today, and that is the natural character of 0:49:52.520 --> 0:49:58.920 continuous speech. These machines always produce speech that sounded stilted, unreal, alien. 0:49:59.000 --> 0:50:01.799 It was never something that would make you feel like 0:50:01.880 --> 0:50:05.200 you were actually being talked to, as much as sort 0:50:05.239 --> 0:50:11.120 of receiving a weird alien code in your language and 0:50:11.160 --> 0:50:14.560 here I just want to read from the stories chapter quote. 0:50:14.760 --> 0:50:18.560 As a result, synthesis often presents itself as an oral 0:50:18.880 --> 0:50:22.800 caricature that can be perceived as an unnatural and sometimes 0:50:22.920 --> 0:50:26.920 amusing rendition of a desired utterance or speech sound. It 0:50:27.000 --> 0:50:31.279 is particularly unique to phonetics and speech science that the 0:50:31.400 --> 0:50:35.759 models used as tools to understand the scientific aspects of 0:50:35.800 --> 0:50:40.000 a complex system produce a signal intended to be heard 0:50:40.200 --> 0:50:43.160 as if it were a human. As such, the quality 0:50:43.160 --> 0:50:47.080 of a speech synthesis can be rather harshly judged because 0:50:47.200 --> 0:50:50.160 the model on which it is based has not accounted 0:50:50.200 --> 0:50:54.239 for the myriad of subtle variations and details that combine 0:50:54.440 --> 0:50:58.640 in natural human speech. So to paraphrase, speech is so 0:50:58.880 --> 0:51:02.279 much more than just the words, And even if you 0:51:02.360 --> 0:51:05.839 can get the words right, there's still something that is 0:51:06.280 --> 0:51:08.319 that is lacking and this is going to take a 0:51:08.400 --> 0:51:10.320 lot of work to try to capture. 0:51:10.640 --> 0:51:13.080 Yeah, this is this is fascinating to think about, you know, 0:51:13.120 --> 0:51:16.640 especially given what you mentioned earlier about it's the importance 0:51:16.800 --> 0:51:21.400 of speech the synthesizer technology to aid people who cannot 0:51:21.560 --> 0:51:24.719 speak or have lost the ability to speak. You know, 0:51:24.760 --> 0:51:27.279 I gave probably one of the most famous, if not 0:51:27.320 --> 0:51:30.279 the most famous examples of this is, of course, the 0:51:30.320 --> 0:51:36.840 speech synthesizer used by theoretical Stephen Hawking. Like one of 0:51:36.640 --> 0:51:39.760 the interesting things about his story with it, as I remember, 0:51:39.880 --> 0:51:42.640 is that just me mentioning it, you can probably sort 0:51:42.640 --> 0:51:45.520 of hear the voice, the synthesized voice of Stephen Hawking 0:51:45.800 --> 0:51:49.120 in your head. And I know that at some point 0:51:49.239 --> 0:51:52.399 like that was you know, an early system he got there, 0:51:52.440 --> 0:51:54.560 and later on in life he had he could have 0:51:54.600 --> 0:51:56.920 switched the voice up, he could have changed the voice, 0:51:57.320 --> 0:52:00.840 and I'm assuming could have maybe improved upon it, but 0:52:00.960 --> 0:52:03.920 by that point he felt that this was his voice. 0:52:03.920 --> 0:52:05.799 You know, you can't switch it up. You know, this 0:52:05.880 --> 0:52:08.200 is this is how I speak, and this is how 0:52:08.239 --> 0:52:11.640 I hear myself. So I always found that that interesting, 0:52:11.680 --> 0:52:13.560 and especially when and then you can compare that to 0:52:13.600 --> 0:52:16.760 some other cases like you know, film credit Roger Ebert 0:52:16.840 --> 0:52:18.800 late in life, you know, you could no longer speak, 0:52:19.080 --> 0:52:21.480 but had I think they had a more robust system 0:52:21.520 --> 0:52:24.839 put together based on samples of you know, the the 0:52:24.880 --> 0:52:29.319 great catalog of his own recorded speeches and reviews and 0:52:29.320 --> 0:52:32.000 so forth that they could draw upon, and then looking 0:52:32.040 --> 0:52:35.200 into the future, you have situations like James Earl Jones's 0:52:35.280 --> 0:52:41.160 Darth Vader voice, that being you know, sort of archived 0:52:41.320 --> 0:52:45.120 and prepared for so that in the future you can 0:52:45.400 --> 0:52:49.480 you can basically have like a machine synthesized version of 0:52:49.480 --> 0:52:52.840 that voice that will stand in as a sort of 0:52:52.920 --> 0:52:55.920 one to one replication of what James Earl Jones did 0:52:55.960 --> 0:52:59.480 in life with the voice acting, or at least so. 0:52:59.480 --> 0:53:01.680 The proponent of the technology would say, I'm sure there 0:53:01.719 --> 0:53:03.480 would be critics who would say it's never going to 0:53:03.520 --> 0:53:04.759 be at one to one. 0:53:05.040 --> 0:53:07.160 Right, right, And then of course there's also the argument 0:53:07.360 --> 0:53:11.480 specifically with only with Darth Vader. Here am I discussing this, 0:53:11.600 --> 0:53:14.719 but obviously the case can be made that like we 0:53:14.840 --> 0:53:19.720 we shouldn't reproduce, you know, deceased actors' voices to continue 0:53:19.719 --> 0:53:24.200 a fictional role. We should employ new living actors. And 0:53:24.280 --> 0:53:27.360 existing living voice actors who can do the voice. I 0:53:27.400 --> 0:53:30.000 think with Darth Vader in particular, you could make a 0:53:30.000 --> 0:53:32.160 strong case for that because there are other voice actors 0:53:32.160 --> 0:53:36.000 who do officially voice act that character and do a 0:53:36.000 --> 0:53:39.719 great job with it. What does it mean if if 0:53:39.719 --> 0:53:44.120 that individual's job is potentially taken by this sort of 0:53:44.239 --> 0:53:48.880 machine likeness of that voice that is authorized based on 0:53:49.239 --> 0:53:53.280 the voice of a you know, of a retired or 0:53:53.400 --> 0:53:55.600 or in some cases you know, deceased individual. 0:53:55.920 --> 0:53:58.160 Well, we're going a little off topic now, but I 0:53:58.200 --> 0:54:00.520 will say that I stand by what I've said before, 0:54:00.520 --> 0:54:03.040 which is I'm firmly in the camp that I prefer 0:54:03.360 --> 0:54:07.600 recasting with a different actor as opposed to using technology 0:54:07.640 --> 0:54:10.560 to try to synthesize the voice or appearance of an 0:54:10.600 --> 0:54:14.279 actor who, for whatever reason cannot be present. People have 0:54:14.280 --> 0:54:17.160 been recasting the same role with different actors for decades. 0:54:17.200 --> 0:54:19.319 That happens all the time. Like, what's the problem with it? 0:54:19.840 --> 0:54:23.400 Yeah? I agree, I agree, But in some cases, is 0:54:23.440 --> 0:54:28.239 it possible that a role that's been established by a 0:54:28.280 --> 0:54:34.439 living actor could not be just masterfully redone by a 0:54:34.640 --> 0:54:38.840 clunky machine with the face of a cherub that is 0:54:39.960 --> 0:54:42.719 manipulated by a sad German Man who needs a haircut. 0:54:42.920 --> 0:54:45.560 I think there's some potential there, like I don't know 0:54:45.640 --> 0:54:46.600 the next James Bond. 0:54:46.640 --> 0:54:50.040 Maybe this is the only film genre I'm interested in 0:54:50.080 --> 0:54:55.239 from now on, high tension espionage movies starring the Euphonia. 0:54:57.040 --> 0:55:00.120 So there you have it. The machine speaks, obviously, I'd 0:55:00.120 --> 0:55:01.400 love to hear from everyone out there if you have 0:55:01.440 --> 0:55:03.480 thoughts on all of this, and certainly anyone out there 0:55:03.480 --> 0:55:08.680 who has direct experience with speech synthesizer technology for one 0:55:08.800 --> 0:55:11.040 use or another right in, we would love to hear 0:55:11.120 --> 0:55:11.480 from you. 0:55:12.080 --> 0:55:17.280 Just a reminder, I just the speech synthesis or speech 0:55:17.360 --> 0:55:21.120 synthesizer is one of the hardest pairs of words to enunciate, 0:55:21.200 --> 0:55:22.680 and I've had to say it so many times in 0:55:22.719 --> 0:55:26.600 this episode. I just want to be recognized, especially for 0:55:26.640 --> 0:55:27.920 the times I probably did it wrong. 0:55:28.719 --> 0:55:31.600 Yes, well it's easy for the babyface machines though, Yeah, 0:55:31.680 --> 0:55:34.279 so at any rate. Yeah, if you want to listen 0:55:34.280 --> 0:55:35.920 to other episodes of Stuff to Blow Your Mind, you 0:55:35.920 --> 0:55:37.239 will find them in the Stuff to Bliw Your Mind 0:55:37.239 --> 0:55:41.160 podcast feed with our core episodes on Tuesdays and Thursdays. Mondays, 0:55:41.160 --> 0:55:43.359 we do a listener mail Wednesdays, we do a short 0:55:43.360 --> 0:55:45.919 form artufactor monster fact, and then on Fridays we set 0:55:45.920 --> 0:55:48.560 aside most serious concerns to just talk about a weird 0:55:48.560 --> 0:55:50.239 film on Weird House Cinema. 0:55:50.400 --> 0:55:54.399 Huge thanks to our excellent audio producer JJ Posway. If 0:55:54.400 --> 0:55:55.799 you would like to get in touch with us with 0:55:55.880 --> 0:55:58.360 feedback on this episode or any other, to suggest a 0:55:58.400 --> 0:56:00.440 topic for the future, or just to say hello, you 0:56:00.480 --> 0:56:03.200 can email us at contact at stuff to Blow your 0:56:03.239 --> 0:56:11.320 Mind dot com. 0:56:11.360 --> 0:56:14.319 Stuff to Blow Your Mind is production of iHeartRadio. For 0:56:14.400 --> 0:56:17.200 more podcasts from my heart Radio, visit the iHeartRadio app, 0:56:17.360 --> 0:56:34.320 Apple Podcasts, or wherever you listen to your favorite shows.