quartz-research-note/content/レイ・カーツワイルのインタビュー(2007).md

[[レイ・カーツワイルのインタビュー日本語訳]]

[Ray Kurzweil \| NAMM.org](https://www.namm.org/library/oral-history/ray-kurzweil)
1
00:00:00,000 --> 00:00:29,980
[Speaker 2] I really appreciate you being here. Thank you very much. Well, it's great to be here. It's great to be at NAMM. I used to come here twice, well, here once a year on the Chicago show. Oh, is that right? Yeah, starting in 83, I think. When we unveiled the Kurzweil 250, we had a prototype yet, it wasn't in production, and we had private showings, and we got a very

2
00:00:29,980 --> 00:00:59,960
[Speaker 1] enthusiastic response, so that was kind of the beginning. Oh, that's neat. Tell me just a little brief history, if you would, on how music became so important to you. Did you have a lot growing up? Well, my father was a famous musician. He was conducting the Bell Symphony, which was the symphony orchestra, the Bell Telephone System. They were on TV a lot. He was head of the Pittsburgh Opera, Mobile Opera, Queens Concert Orchestra,

3
00:00:59,960 --> 00:01:29,940
[Speaker 2] founded the music department at the Queensport College, and composer. So he taught me piano when I was six. Actually, in the 60s, he became very excited about the Moog synthesizer and switched on Bach. And I had an interest in computers at that time. I'd been building computers. And he said, you know, someday you're going to combine computers and music, and do something with computer music, which really didn't exist back then, because the first introduction

4
00:01:29,940 --> 00:01:59,920
[Speaker 2] to combination of combination of technology and music was analog synthesis, which did take him, did pique his interest. He died in 1970, but I've always kept that sort of challenge in mind. My primary interest is pattern recognition as part of artificial intelligence, trying to emulate the pattern recognition capability of human beings. And I actually did a

5
00:01:59,920 --> 00:02:17,910
[Speaker 2] project in the project in the 70s. I created the first Omni font, any type font character recognition. And then happened to sit next to a blind gentleman on a plane who said, blindness is not really a handicap. He can do anything. He's traveling around the world, but he can't really read ordinary print. And that was a handicap.

6
00:02:17,910 --> 00:02:24,410
[Speaker 2] And so I then devoted that character recognition technology to the blind reading problem because

7
00:02:24,410 --> 00:02:27,860
[Speaker 2] it was sort of a solution in search of a problem.

8
00:02:27,860 --> 00:02:29,970
[Speaker 2] And we had to create two other technologies.

9
00:02:29,970 --> 00:02:33,910
[Speaker 2] We created the first flatbed scanner and the first speech synthesis.

10
00:02:33,910 --> 00:02:41,020
[Speaker 2] And we combined those three technologies, which today are ubiquitous, character recognition,

11
00:02:41,020 --> 00:02:45,520
[Speaker 2] flatbed scanning, and full text-to-speech synthesis and created the first print-to-speech

12
00:02:45,520 --> 00:02:48,240
[Speaker 2] reading machine for the blind.

13
00:02:48,240 --> 00:02:51,900
[Speaker 2] And we introduced it.

14
00:02:51,900 --> 00:02:53,760
[Speaker 2] It was on all three nightly networks.

15
00:02:53,760 --> 00:02:59,360
[Speaker 2] Walter Crockett used it for his signature sign-off, and that's the way it was, January 13, 1976.

16
00:02:59,360 --> 00:03:06,120
[Speaker 2] A few days later, I was on the Today Show, and Stevie Wonder happened to catch that show.

17
00:03:06,120 --> 00:03:10,420
[Speaker 2] And he called us up.

18
00:03:10,420 --> 00:03:13,120
[Speaker 2] Our receptionist didn't really believe it was him, but put him through anyway.

19
00:03:13,120 --> 00:03:15,840
[Speaker 2] Anyway, it said, "Well, this is amazing.

20
00:03:15,840 --> 00:03:20,060
[Speaker 2] I have to stop by, and I want to buy one."

21
00:03:20,060 --> 00:03:23,980
[Speaker 2] We actually didn't have one, but we very quickly finished one up.

22
00:03:23,980 --> 00:03:25,360
[Speaker 2] He came by.

23
00:03:25,360 --> 00:03:26,840
[Speaker 2] We spent a few hours showing him how to use it.

24
00:03:26,840 --> 00:03:31,840
[Speaker 2] He went off in a taxi with his reading machine, Kurzweil reading machine.

25
00:03:31,840 --> 00:03:34,420
[Speaker 2] It was our first customer.

26
00:03:34,420 --> 00:03:39,860
[Speaker 2] But that was 1976, and that started a relationship which has continued to this day.

27
00:03:39,860 --> 00:03:42,660
[Speaker 2] So that's 30 years now.

28
00:03:42,660 --> 00:03:46,280
[Speaker 2] And so over the years we had a lot of discussions.

29
00:03:46,280 --> 00:03:54,660
[Speaker 2] He's actually pretty savvy about technology, both because it's a great equalizer for disabilities,

30
00:03:54,660 --> 00:03:58,680
[Speaker 2] and because technology plays such an important role in music.

31
00:03:58,680 --> 00:04:05,500
[Speaker 2] And in 1982 he was showing me around a new studio he had called Wonderland here in Los Angeles.

32
00:04:05,500 --> 00:04:10,300
[Speaker 2] And it was lamenting the state of affairs that there was really these two disconnected worlds

33
00:04:10,300 --> 00:04:11,440
[Speaker 2] of musical instruments.

34
00:04:11,440 --> 00:04:18,260
[Speaker 2] There was the electronic world where you could do, you had all these tremendous control capabilities.

35
00:04:18,260 --> 00:04:23,240
[Speaker 2] You could play one part and the computer would remember it, and then you could play it back

36
00:04:23,240 --> 00:04:29,620
[Speaker 2] from the computer's memory and play another line over it, and you could build up multi-line orchestrations.

37
00:04:29,620 --> 00:04:32,040
[Speaker 2] You could modify the sounds.

38
00:04:32,040 --> 00:04:39,080
[Speaker 2] But the sounds you had to work with in that electronic world were very thin, synthetic sounding sounds.

39
00:04:39,080 --> 00:04:43,520
[Speaker 2] And then there was these sounds of 19th century acoustic instruments like the piano, guitar, violin.

40
00:04:43,520 --> 00:04:51,260
[Speaker 2] And these were still the sounds of choice of musicians because they had a lot of deep, complex resonance.

41
00:04:51,260 --> 00:04:54,380
[Speaker 2] But they're very hard to play.

42
00:04:54,380 --> 00:04:58,360
[Speaker 2] I mean most musicians can't play most instruments.

43
00:04:58,360 --> 00:05:02,390
[Speaker 2] And even if you're a virtuoso and can play them all, you can't play them simultaneously.

44
00:05:02,390 --> 00:05:03,980
[Speaker 2] Most of you can only play one note at a time.

45
00:05:03,980 --> 00:05:08,440
[Speaker 2] So you can't modify the sounds except maybe a few modification techniques.

46
00:05:08,440 --> 00:05:12,240
[Speaker 2] You can do vibrato and a violin, but it's very limited.

47
00:05:12,240 --> 00:05:17,640
[Speaker 2] So wouldn't it be great if we could take this very rich array of control techniques that you

48
00:05:17,640 --> 00:05:22,700
[Speaker 2] have in the electronic world and apply it to these very rich, complex sounds of choice of the

49
00:05:22,700 --> 00:05:24,320
[Speaker 2] acoustic world.

50
00:05:24,320 --> 00:05:27,600
[Speaker 2] And then maybe you could actually create new sounds that have the complexity of acoustic

51
00:05:27,600 --> 00:05:30,960
[Speaker 2] sounds but aren't sounds that any acoustic instrument could make.

52
00:05:30,960 --> 00:05:36,440
[Speaker 2] You could open up a whole new world of synthetic sounds that weren't so simple and synthetic and

53
00:05:36,440 --> 00:05:41,660
[Speaker 2] thin as the electronic world was used to at that time.

54
00:05:41,660 --> 00:05:45,660
[Speaker 2] And I felt actually using advanced signal processing and some pattern recognition insights that

55
00:05:45,660 --> 00:05:48,060
[Speaker 2] we could do that.

56
00:05:48,060 --> 00:05:52,780
[Speaker 2] And so we started Kurzweil Music then in 1982, July 1st actually.

57
00:05:52,780 --> 00:05:56,400
[Speaker 2] C.V. Wonder was our advisor.

58
00:05:56,400 --> 00:06:01,640
[Speaker 2] We set to create the Kurzweil 250, the first instrument that could really recreate the acoustic

59
00:06:01,640 --> 00:06:03,300
[Speaker 2] piano.

60
00:06:03,300 --> 00:06:05,440
[Speaker 2] And there are a lot of challenges.

61
00:06:05,440 --> 00:06:08,660
[Speaker 2] You might think that, well, you just sample the piano and it'll sound like a piano.

62
00:06:08,660 --> 00:06:13,660
[Speaker 2] But samplers, particularly at that time, for example, would loop the last waveform and then

63
00:06:13,660 --> 00:06:16,220
[Speaker 2] have a decaying envelope.

64
00:06:16,220 --> 00:06:21,220
[Speaker 2] That doesn't work for the piano because the overtones in a piano are not perfect multiples of the fundamental.

65
00:06:21,220 --> 00:06:23,220
[Speaker 2] They're a little off.

66
00:06:23,220 --> 00:06:24,220
[Speaker 2] They're called anharmonic.

67
00:06:24,220 --> 00:06:26,220
[Speaker 2] That gives the piano its sort of rich character.

68
00:06:26,220 --> 00:06:29,720
[Speaker 2] Well, if you loop one waveform, then all the overtones become perfect multiples.

69
00:06:29,720 --> 00:06:34,100
[Speaker 2] It sounds like an organ tone and it loses its piano character.

70
00:06:34,100 --> 00:06:36,780
[Speaker 2] So that's an insight from pattern recognition.

71
00:06:36,780 --> 00:06:41,460
[Speaker 2] But that's very challenging for sampling, particularly then when memory was expensive because you can't

72
00:06:41,460 --> 00:06:47,640
[Speaker 2] really afford to record the entire 20-second decay of a piano note.

73
00:06:47,640 --> 00:06:50,640
[Speaker 2] If you hit a middle C harder, it's not just louder.

74
00:06:50,640 --> 00:06:57,840
[Speaker 2] It has a whole different time-bearing pitch contour.

75
00:06:57,840 --> 00:07:03,820
[Speaker 2] The different volume levels have a completely different time-bearing timbre.

76
00:07:03,820 --> 00:07:06,880
[Speaker 2] And you can't really capture all of that with sampling.

77
00:07:06,880 --> 00:07:11,640
[Speaker 2] And if you try to bend the pitch too much, it changes the character of the tones in unrealistic

78
00:07:11,640 --> 00:07:12,640
[Speaker 2] ways.

79
00:07:12,640 --> 00:07:15,980
[Speaker 2] The tones won't interact with each other when you put the pedal down.

80
00:07:15,980 --> 00:07:21,400
[Speaker 2] So a lot of complexities of real-life piano that you really can't capture with sampling unless

81
00:07:21,400 --> 00:07:25,440
[Speaker 2] you add some other elements of signal processing and pattern recognition.

82
00:07:25,440 --> 00:07:26,440
[Speaker 2] So that's what we sought to do.

83
00:07:26,440 --> 00:07:37,440
[Speaker 2] And we came actually to my first NAMM show, which was here in Anaheim in '83, and showed

84
00:07:37,440 --> 00:07:39,440
[Speaker 2] our prototype of the Kurzweil 250.

85
00:07:39,440 --> 00:07:43,780
[Speaker 2] And people were pretty much blown away because it really did sound like a piano.

86
00:07:43,780 --> 00:07:45,440
[Speaker 2] We started shipping it in '84.

87
00:07:45,440 --> 00:07:52,780
[Speaker 2] And it did get recognized as the first electronic instrument that could recreate the piano, which

88
00:07:52,780 --> 00:07:54,280
[Speaker 2] really is the most challenging.

89
00:07:54,280 --> 00:07:57,440
[Speaker 2] But it also could do other orchestral instruments.

90
00:07:57,440 --> 00:08:02,500
[Speaker 2] And it has maintained its sort of edge in terms of sound quality and realism, particularly

91
00:08:02,500 --> 00:08:04,540
[Speaker 2] of acoustic instruments.

92
00:08:04,540 --> 00:08:07,880
[Speaker 2] But sound quality in general.

93
00:08:07,880 --> 00:08:09,440
[Speaker 2] So that's how we got started.

94
00:08:09,440 --> 00:08:16,440
[Speaker 1] You know, I interviewed a former president of Baldwin, and I asked him, you know, was there

95
00:08:16,440 --> 00:08:21,440
[Speaker 1] any point during his career that they worried about the synthesizer replacing the piano?

96
00:08:21,440 --> 00:08:24,440
[Speaker 1] And he says, "Not until the Kurzweil came out.

97
00:08:24,440 --> 00:08:25,440
[Speaker 1] Then we started worrying."

98
00:08:25,440 --> 00:08:26,440
[Speaker 1] Yeah.

99
00:08:26,440 --> 00:08:33,440
[Speaker 2] Well, I think, you know, digital pianos for the home market have significantly cut into the,

100
00:08:33,440 --> 00:08:35,440
[Speaker 2] particularly the low-end market.

101
00:08:35,440 --> 00:08:39,440
[Speaker 2] You know, parents that want to buy a piano for their eight-year-old child.

102
00:08:39,440 --> 00:08:42,440
[Speaker 2] Because there are a lot of advantages to electronic piano.

103
00:08:42,440 --> 00:08:49,440
[Speaker 2] For the same price range, you can get a better quality sound in a digital piano than an acoustic.

104
00:08:49,440 --> 00:08:57,440
[Speaker 2] You know, at the high end, there's still a market and still an advantage for acoustic pianos.

105
00:08:57,440 --> 00:09:04,440
[Speaker 2] But if you're talking about just a routine piano for Sally to take piano lessons, you get a

106
00:09:04,440 --> 00:09:05,440
[Speaker 2] lot of value.

107
00:09:05,440 --> 00:09:07,440
[Speaker 2] And then there's a lot of other capabilities, you know.

108
00:09:07,440 --> 00:09:12,440
[Speaker 2] When Sally learns how to play a piece, she can also play it on the violin and the human

109
00:09:12,440 --> 00:09:14,440
[Speaker 2] voice and all these other instruments.

110
00:09:14,440 --> 00:09:24,440
[Speaker 2] And then electronic instruments have autoplay and can help teach you how to play the piano,

111
00:09:24,440 --> 00:09:27,440
[Speaker 2] can record what you're playing with sequencers.

112
00:09:27,440 --> 00:09:30,440
[Speaker 2] So you have none of those capabilities in an acoustic piano.

113
00:09:30,440 --> 00:09:32,440
[Speaker 2] You don't have to tune them.

114
00:09:32,440 --> 00:09:38,440
[Speaker 2] You can use headphones so people can practice without disturbing other people.

115
00:09:38,440 --> 00:09:39,440
[Speaker 2] So a lot of advantages.

116
00:09:39,440 --> 00:09:45,440
[Speaker 2] And it has really cut into the, I'd say, the upright piano market.

117
00:09:45,440 --> 00:09:46,440
[Speaker 2] Interesting.

118
00:09:46,440 --> 00:09:54,440
[Speaker 1] And one of the things that I think that you guys set out to do was not just a piano, but

119
00:09:54,440 --> 00:09:58,440
[Speaker 1] what other musical instruments can we recreate?

120
00:09:58,440 --> 00:10:04,440
[Speaker 2] Well, the piano is the most challenging, but it really can recreate any orchestral instrument.

121
00:10:04,440 --> 00:10:10,440
[Speaker 2] And also can then break down these instruments into their components and create synthetic sounds.

122
00:10:10,440 --> 00:10:16,440
[Speaker 2] That is, you know, a new sound that no acoustic instrument could create but has the complexity

123
00:10:16,440 --> 00:10:18,440
[Speaker 2] and richness of an acoustic sound.

124
00:10:18,440 --> 00:10:23,440
[Speaker 2] Maybe because you started with an acoustic sound and modified it to be unrecognizable, but

125
00:10:23,440 --> 00:10:25,440
[Speaker 2] it nonetheless keeps its complex character.

126
00:10:25,440 --> 00:10:36,440
[Speaker 2] So, you know, if you just start from the ground up doing, say, analog synthesis, that was a very groundbreaking

127
00:10:36,440 --> 00:10:39,440
[Speaker 2] development when it occurred and it was a new class of sounds.

128
00:10:39,440 --> 00:10:45,440
[Speaker 2] But there's a limited complexity you can create by just building up oscillators.

129
00:10:45,440 --> 00:10:50,440
[Speaker 2] By starting with the richness and complexity of acoustic sounds and then modifying it using

130
00:10:50,440 --> 00:10:56,440
[Speaker 2] a whole panoply of signal processing methods, you can, you know, keep the complexity and the

131
00:10:56,440 --> 00:11:00,440
[Speaker 2] musical depth but create a whole new class of sounds.

132
00:11:00,440 --> 00:11:03,440
[Speaker 2] And then you can modify it with all kinds of other synthesis techniques.

133
00:11:03,440 --> 00:11:12,440
[Speaker 2] And so we have a new chip now that does a significant amount of digital signal processing on each channel.

134
00:11:12,440 --> 00:11:17,440
[Speaker 2] So you could start with the sampled sound but then modify it by putting it basically through

135
00:11:17,440 --> 00:11:27,440
[Speaker 2] a whole complex set of synthesizer sound modification capabilities per channel and then apply more

136
00:11:27,440 --> 00:11:30,440
[Speaker 2] signal processing to the mix sound.

137
00:11:30,440 --> 00:11:33,440
[Speaker 2] So there are a lot of capabilities.

138
00:11:33,440 --> 00:11:37,440
[Speaker 2] And one of the things, I mean, another whole interest I have is in tracking technology trends.

139
00:11:37,440 --> 00:11:43,440
[Speaker 2] And I did that because of my interest in being an inventor and because I realized timing was critical

140
00:11:43,440 --> 00:11:46,440
[Speaker 2] and most inventions fail because the timing is wrong.

141
00:11:46,440 --> 00:11:50,440
[Speaker 2] And so I developed these models of how technology evolves.

142
00:11:50,440 --> 00:11:53,440
[Speaker 2] And it evolves actually in very predictable ways.

143
00:11:53,440 --> 00:12:06,440
[Speaker 2] I have a theory called the law of accelerating returns that indicates, that says that information technology in many different areas is basically doubling its power every year in terms of price performance.

144
00:12:06,440 --> 00:12:07,440
[Speaker 2] And capacity.

145
00:12:07,440 --> 00:12:11,440
[Speaker 2] Doubling every year is very phenomenal exponential growth.

146
00:12:11,440 --> 00:12:13,440
[Speaker 2] That's a factor of a thousand in ten years.

147
00:12:13,440 --> 00:12:15,440
[Speaker 2] A million in 20 years.

148
00:12:15,440 --> 00:12:18,440
[Speaker 2] A billion in 30 years.

149
00:12:18,440 --> 00:12:23,440
[Speaker 2] So it's been over 20 years since we introduced the Curso 250.

150
00:12:23,440 --> 00:12:31,440
[Speaker 2] So in those 20 years, information technology, computer technology, digital signal processing, it's all gotten a million times more powerful.

151
00:12:31,440 --> 00:12:45,440
[Speaker 2] So, you know, we can now do in a low end instrument, you know, thousands of times more transformations in capability than was feasible in an expensive instrument 20 years ago.

152
00:12:45,440 --> 00:12:46,440
[Speaker 2] So, you know, that's going to continue.

153
00:12:46,440 --> 00:12:52,440
[Speaker 2] You know, because I have this whole new phenomenon of software synthesis where with a PC you can do some pretty impressive things.

154
00:12:52,440 --> 00:12:59,440
[Speaker 1] How long did you stay focused in the company as far as developing products?

155
00:12:59,440 --> 00:13:03,440
[Speaker 2] Well, I started the company in '82.

156
00:13:03,440 --> 00:13:05,440
[Speaker 2] We introduced Curso 250 in '84.

157
00:13:05,440 --> 00:13:10,440
[Speaker 2] We had our first chip based products, Curso 1000, a few years after that.

158
00:13:10,440 --> 00:13:19,440
[Speaker 2] We sold the company to Yong Chang, a Korean piano manufacturer in 1990.

159
00:13:19,440 --> 00:13:24,440
[Speaker 2] And I remained actively involved through '95.

160
00:13:24,440 --> 00:13:34,440
[Speaker 2] After that I was not directly involved, although I stayed in close touch with all the engineers through the sort of ups and downs.

161
00:13:34,440 --> 00:13:44,440
[Speaker 2] A number of key people, including myself, left around '95, '96, '97 in Yong Chang got into some financial difficulties in recent years.

162
00:13:44,440 --> 00:13:49,440
[Speaker 2] And we were actually concerned how to revitalize the company.

163
00:13:49,440 --> 00:14:00,440
[Speaker 2] I think that's worked out actually quite well because Hyundai has bought the company and they have very substantial resources and they know how to manufacture high quality products.

164
00:14:00,440 --> 00:14:02,440
[Speaker 2] So they're revitalizing the company.

165
00:14:02,440 --> 00:14:09,440
[Speaker 2] I think it's going to be a strong combination with their resources and the Kurzweil brand and the Kurzweil core technology.

166
00:14:09,440 --> 00:14:21,440
[Speaker 2] But the technology stayed intact and there was one key project, the Mara chip, which is some products being introduced here at NAMM are based on that, which remained.

167
00:14:21,440 --> 00:14:26,440
[Speaker 2] And it's really a very cutting edge chip.

168
00:14:26,440 --> 00:14:29,440
[Speaker 2] So the core technology remains very strong.

169
00:14:29,440 --> 00:14:33,440
[Speaker 2] There's very strong sound where the key engineers have remained throughout.

170
00:14:33,440 --> 00:14:37,440
[Speaker 2] Hyundai is now building up the company again.

171
00:14:37,440 --> 00:14:40,440
[Speaker 2] Our research and development has quadrupled in the last year.

172
00:14:40,440 --> 00:14:45,440
[Speaker 1] So back up a little second and ask you an off the wall question.

173
00:14:45,440 --> 00:14:58,440
[Speaker 2] Oh, and I'm actually back now advising Kurzweil Music again, helping them with technology strategy and also being a spokesperson as the founder.

174
00:14:58,440 --> 00:15:03,440
[Speaker 2] But yeah, I'm involved again pretty closely with the new team.

175
00:15:03,440 --> 00:15:04,440
[Speaker 1] Excellent.

176
00:15:04,440 --> 00:15:07,440
[Speaker 1] So is there going to be new products?

177
00:15:07,440 --> 00:15:25,440
[Speaker 2] Yeah, there's new products here, MP3, which is very impressive, 128 voices, highest, 32-bit voices, 128 of them.

178
00:15:25,440 --> 00:15:31,440
[Speaker 2] Each voice has significant amount of digital signal processing capability per voice.

179
00:15:31,440 --> 00:15:36,440
[Speaker 2] And then there's a very sophisticated effects processor for the mixed output.

180
00:15:36,440 --> 00:15:42,440
[Speaker 2] And lots of other features, advanced sequencers and sound modification.

181
00:15:42,440 --> 00:15:52,440
[Speaker 2] So it's a pretty high-end instrument, but it's going to be, I mean, list price of $2,500, so street price will be less than that.

182
00:15:52,440 --> 00:15:54,440
[Speaker 2] It's actually quite impressive.

183
00:15:54,440 --> 00:16:01,440
[Speaker 2] And then there's a SP2, Stage Piano 2, which is kind of a lower-end version of that, but still very impressive.

184
00:16:01,440 --> 00:16:04,440
[Speaker 2] Sixty-four voices, which is still a lot.

185
00:16:04,440 --> 00:16:08,440
[Speaker 2] And that'll be something like $1,500 at this price.

186
00:16:08,440 --> 00:16:11,440
[Speaker 2] So those are pretty impressive products and they're here.

187
00:16:11,440 --> 00:16:15,440
[Speaker 2] In fact, Stevie Wonder was just playing them a couple hours ago.

188
00:16:15,440 --> 00:16:16,440
[Speaker 2] How great.

189
00:16:16,440 --> 00:16:25,440
[Speaker 1] I was going to ask you, what's your, who's your thought of the father of electronic music?

190
00:16:25,440 --> 00:16:32,440
[Speaker 1] There's a lot of names that float around in my head, and I wonder who you would consider that to be.

191
00:16:32,440 --> 00:16:36,440
[Speaker 2] Well, I mean, Bob Moog really put it on the map.

192
00:16:36,440 --> 00:16:39,440
[Speaker 2] He was inspired by theremin.

193
00:16:39,440 --> 00:16:46,440
[Speaker 2] In fact, he started building theremins at a young age, but then took it in a whole new direction

194
00:16:46,440 --> 00:16:55,440
[Speaker 2] and established the basis of synthesis using the technology of that time, which was analog synthesis and oscillators.

195
00:16:55,440 --> 00:16:59,440
[Speaker 2] He created a whole new class of instrument and a whole new class of sound.

196
00:16:59,440 --> 00:17:02,440
[Speaker 2] He created a lot of excitement.

197
00:17:02,440 --> 00:17:11,440
[Speaker 2] I mean, switched on Bach, that album by Walter Carlos, I think he was Walter at that time, created a big buzz.

198
00:17:11,440 --> 00:17:18,440
[Speaker 2] So my father, who was a classical musician, opera conductor and timpani conductor and concert pianist,

199
00:17:18,440 --> 00:17:22,440
[Speaker 2] got very excited about that.

200
00:17:22,440 --> 00:17:26,440
[Speaker 2] And it attracted tremendous interest from the classical world to the pop world.

201
00:17:26,440 --> 00:17:31,440
[Speaker 2] And really put, I mean, created the synthesizer.

202
00:17:31,440 --> 00:17:33,440
[Speaker 2] So there were roots to it.

203
00:17:33,440 --> 00:17:39,440
[Speaker 2] He didn't create it out of nothing, but that really was the beginning of the synthesizer.

204
00:17:39,440 --> 00:17:46,440
[Speaker 2] And then there were various efforts for digital synthesizers and samplers.

205
00:17:46,440 --> 00:17:48,440
[Speaker 2] Emu played an important role.

206
00:17:48,440 --> 00:17:58,440
[Speaker 2] And our goal was actually to go beyond just sampling to bring some signal processing sophistication

207
00:17:58,440 --> 00:18:05,440
[Speaker 2] to really capture some of these complex effects of acoustic instruments that you can't just get by sampling.

208
00:18:05,440 --> 00:18:08,440
[Speaker 2] Because samplers were pretty unsuccessful in capturing the piano.

209
00:18:08,440 --> 00:18:14,440
[Speaker 1] We're definitely going to have to do a part two one of these days, because I'm running out of time.

210
00:18:14,440 --> 00:18:23,440
[Speaker 1] But I do want to ask you, do you have the tape of your 1965 television appearance?

211
00:18:23,440 --> 00:18:25,440
[Speaker 1] Is that I got a secret?

212
00:18:25,440 --> 00:18:26,440
[Speaker 2] Yeah.

213
00:18:26,440 --> 00:18:33,440
[Speaker 2] In fact, I was shown recently, I did a three hour interview on a program called In Depth on Book TV, C-SPAN 2.

214
00:18:33,440 --> 00:18:37,440
[Speaker 2] I started out the program by displaying that.

215
00:18:37,440 --> 00:18:39,440
[Speaker 2] We can send you a copy.

216
00:18:39,440 --> 00:18:41,440
[Speaker 1] I'd love to see it, yeah.

217
00:18:41,440 --> 00:18:44,440
[Speaker 2] But yeah, actually, that was actually a music project I did in high school,

218
00:18:44,440 --> 00:18:50,440
[Speaker 2] where I used pattern recognition to find patterns in melodies.

219
00:18:50,440 --> 00:18:54,440
[Speaker 2] So I'd feed in Bach melodies or Chopin or Beethoven,

220
00:18:54,440 --> 00:18:58,440
[Speaker 2] and it would actually model the types of patterns that those composers used,

221
00:18:58,440 --> 00:19:01,440
[Speaker 2] and then compose original music using those patterns.

222
00:19:01,440 --> 00:19:07,440
[Speaker 2] And it would sound like a second-rate student of Mozart or Chopin, as the case may be.

223
00:19:07,440 --> 00:19:13,440
[Speaker 2] And I won some science contests, got to meet President Johnson.

224
00:19:13,440 --> 00:19:17,440
[Speaker 2] I was invited on this network show.

225
00:19:17,440 --> 00:19:21,440
[Speaker 2] There weren't very many TV shows back then, so I've got a secret.

226
00:19:21,440 --> 00:19:23,440
[Speaker 2] And my secret, I came on, I played a piece of music,

227
00:19:23,440 --> 00:19:26,440
[Speaker 2] and my secret was that I built a computer that composed the music.

228
00:19:26,440 --> 00:19:32,440
[Speaker 1] Did most Americans even know what a computer was then, you think?

229
00:19:32,440 --> 00:19:36,440
[Speaker 2] Yeah, there had been publicity about these giant brains,

230
00:19:36,440 --> 00:19:38,440
[Speaker 2] and there was already speculation about, you know,

231
00:19:38,440 --> 00:19:41,440
[Speaker 2] are they thinking and what will they be able to do?

232
00:19:41,440 --> 00:19:44,440
[Speaker 2] And some of these early computers were able to do things

233
00:19:44,440 --> 00:19:48,440
[Speaker 2] that professional mathematicians had not been able to do.

234
00:19:48,440 --> 00:19:56,440
[Speaker 2] Interestingly, ironically, the history of artificial intelligence is the opposite of human skill.

235
00:19:56,440 --> 00:20:00,440
[Speaker 2] Computers very quickly learned how to do things that professional humans do,

236
00:20:00,440 --> 00:20:04,440
[Speaker 2] like solving mathematical theorems, diagnosing disease.

237
00:20:04,440 --> 00:20:09,440
[Speaker 2] But today they still struggle, say, telling the difference between a dog and a cat,

238
00:20:09,440 --> 00:20:11,440
[Speaker 2] and things that a five-year-old child can do.

239
00:20:11,440 --> 00:20:15,440
[Speaker 2] They're actually learning how to tell a different machine, a dog and a cat.

240
00:20:15,440 --> 00:20:16,440
[Speaker 2] But it's been backwards.

241
00:20:16,440 --> 00:20:18,440
[Speaker 2] They first learned to do what adults can do,

242
00:20:18,440 --> 00:20:21,440
[Speaker 2] and now they're backing up and learning what children can do.

243
00:20:21,440 --> 00:20:27,440
[Speaker 2] But there had been a lot of publicity even in the '50s about these giant brains

244
00:20:27,440 --> 00:20:30,440
[Speaker 2] and what they will eventually be able to do.

245
00:20:30,440 --> 00:20:33,440
[Speaker 2] And there was already perception of this accelerating returns.

246
00:20:33,440 --> 00:20:36,440
[Speaker 2] Computers were getting twice as powerful every year.

247
00:20:36,440 --> 00:20:39,440
[Speaker 2] So yes, people had heard of computers.

248
00:20:39,440 --> 00:20:43,440
[Speaker 2] But they weren't using them because when I started using computers,

249
00:20:43,440 --> 00:20:48,440
[Speaker 2] which was about 1960, there were like 10 or 12 computers in New York.

250
00:20:48,440 --> 00:20:51,440
[Speaker 2] And one of which I had access to.

251
00:20:51,440 --> 00:20:56,440
[Speaker 1] Well, you're a good cat to spend some time with me.

252
00:20:56,440 --> 00:20:57,440
[Speaker 1] I really do appreciate it.

253
00:20:57,440 --> 00:20:58,440
[Speaker 1] My pleasure.

254
00:20:58,440 --> 00:20:59,440
[Speaker 2] Thank you very much.

255
00:20:59,440 --> 00:21:00,440
[Speaker 1] Thank you very much.