WEBVTT - Can AI Compose Good Music?

0:00:00.760 --> 0:00:04.600
<v Speaker 1>Earlier this month, the Prague Philharmonic Orchestra got on stage

0:00:04.640 --> 0:00:07.360
<v Speaker 1>to play a symphony written by the famous check composer

0:00:07.400 --> 0:00:11.240
<v Speaker 1>Antony Davor Jack. The orchestra was premiering a piece of

0:00:11.320 --> 0:00:27.040
<v Speaker 1>music the world had never heard before. Divorjack died more

0:00:27.080 --> 0:00:29.760
<v Speaker 1>than a hundred years ago, and he left behind just

0:00:29.960 --> 0:00:33.479
<v Speaker 1>the beginnings of a composition, just two sheets of music.

0:00:34.640 --> 0:00:38.200
<v Speaker 1>So a computer program powered by AI studied the rest

0:00:38.240 --> 0:00:42.640
<v Speaker 1>of divor Jack's music and completed the composer's unfinished work.

0:00:43.960 --> 0:00:46.920
<v Speaker 1>This is what the software spit out, something in the

0:00:46.960 --> 0:00:51.240
<v Speaker 1>style of Davor Jack that was still an entirely new symphony.

0:00:53.120 --> 0:00:56.520
<v Speaker 1>For years, we've worried about robots replacing the jobs of

0:00:56.560 --> 0:01:01.120
<v Speaker 1>truck drivers and call center asians and accountants, but experts

0:01:01.120 --> 0:01:04.600
<v Speaker 1>said creative jobs were going to be safe, that computers

0:01:04.640 --> 0:01:07.319
<v Speaker 1>were still so far from making things that are new

0:01:07.360 --> 0:01:11.720
<v Speaker 1>and subjective that move us. So consider the fact that

0:01:11.800 --> 0:01:32.200
<v Speaker 1>AI wrote most of this. Today, in the show Reporter,

0:01:32.280 --> 0:01:36.480
<v Speaker 1>in the tallya Drosiac visits three musicians using artificial intelligence

0:01:36.520 --> 0:01:40.000
<v Speaker 1>to make their music, including the guy behind this robo

0:01:40.080 --> 0:01:45.000
<v Speaker 1>Divorgac symphony. If computers can now compose something this beautiful,

0:01:45.520 --> 0:01:49.280
<v Speaker 1>what's left for us humans to do. Am Ito, you're

0:01:49.320 --> 0:02:19.440
<v Speaker 1>listening to Decrypted stay with us. M Hey Nat, welcome

0:02:19.440 --> 0:02:22.680
<v Speaker 1>to the show. Thanks for having me. So you're our

0:02:22.760 --> 0:02:25.679
<v Speaker 1>European tech reporter out of Brussels, and you normally write

0:02:25.720 --> 0:02:30.200
<v Speaker 1>about Europe's regulatory crackdown on the tech industry, but today

0:02:30.240 --> 0:02:33.840
<v Speaker 1>we're talking about something completely different. Yep. So we've been

0:02:33.840 --> 0:02:36.720
<v Speaker 1>writing about the use of AI and self driving cars

0:02:36.760 --> 0:02:40.680
<v Speaker 1>and chatbots and voice assistants, but in the world of music,

0:02:40.720 --> 0:02:43.840
<v Speaker 1>this is all still kind of new. I mean, musicians

0:02:43.960 --> 0:02:46.480
<v Speaker 1>just started using AI as a tool a few years ago.

0:02:49.040 --> 0:02:51.880
<v Speaker 1>So I talked to three different composers who are using

0:02:51.919 --> 0:02:55.040
<v Speaker 1>AI in really different ways. The first guy I want

0:02:55.040 --> 0:02:59.079
<v Speaker 1>to introduce you to is Ben Waki. He's a French musician.

0:02:59.680 --> 0:03:02.600
<v Speaker 1>He's known in France for these French pop songs, but

0:03:02.639 --> 0:03:06.080
<v Speaker 1>he's also composed music for artists like the famous French

0:03:06.160 --> 0:03:09.480
<v Speaker 1>rocker Johnny Halliday. So this is a pretty famous guy

0:03:09.520 --> 0:03:12.720
<v Speaker 1>out of France. Yeah, he's well known, but he's gotten

0:03:12.720 --> 0:03:16.480
<v Speaker 1>even more attention recently because of his experimentations with AI

0:03:16.520 --> 0:03:20.040
<v Speaker 1>generated music. And how did he get into that. So

0:03:20.160 --> 0:03:23.320
<v Speaker 1>there's this guy called Francois Pachet who for a long

0:03:23.360 --> 0:03:26.680
<v Speaker 1>time was ahead of the research lab run by Sony

0:03:26.760 --> 0:03:30.560
<v Speaker 1>in Paris, and Paschet has kind of become known as

0:03:30.680 --> 0:03:34.520
<v Speaker 1>the godfather of AI music because he's really done a

0:03:34.520 --> 0:03:38.040
<v Speaker 1>lot of research in this area. France called me a

0:03:38.040 --> 0:03:42.320
<v Speaker 1>long time ago. He discovered my my songs in the

0:03:42.400 --> 0:03:46.200
<v Speaker 1>late nineties, and he was interested by my way of

0:03:46.400 --> 0:03:52.840
<v Speaker 1>composing songs, always searching for unexpected coat changes, and and

0:03:52.920 --> 0:03:55.360
<v Speaker 1>he invited ben Wata's lab to try out some of

0:03:55.400 --> 0:03:57.920
<v Speaker 1>the tools that he's been developing over the past few years.

0:03:58.720 --> 0:04:01.720
<v Speaker 1>And one of the tools that Pesh's team developed is

0:04:01.760 --> 0:04:05.000
<v Speaker 1>this machine learning tool that generates music. So how does

0:04:05.040 --> 0:04:09.920
<v Speaker 1>it work? So first, the program basically ingests a bunch

0:04:09.960 --> 0:04:12.560
<v Speaker 1>of sheets of music that the musician wants the program

0:04:12.600 --> 0:04:16.240
<v Speaker 1>to train on, and then the computer cuts it up

0:04:16.279 --> 0:04:19.240
<v Speaker 1>into really tiny bits of music and rearranges it into

0:04:19.279 --> 0:04:22.520
<v Speaker 1>a whole new composition. So let me give you an example.

0:04:22.680 --> 0:04:25.880
<v Speaker 1>So a few years ago, ben wa fed this program

0:04:25.920 --> 0:04:29.440
<v Speaker 1>four seventy lead sheets of jazz standards from the thirties

0:04:29.600 --> 0:04:33.120
<v Speaker 1>up through the fifties and sixties, and out came these

0:04:33.120 --> 0:04:42.679
<v Speaker 1>two bars of music that he really liked, very jazzy.

0:04:43.000 --> 0:04:46.920
<v Speaker 1>I can definitely sense the influence. Yeah, so Ben Wah

0:04:47.000 --> 0:04:50.279
<v Speaker 1>really liked the sound of what the computer came up with.

0:04:50.760 --> 0:04:56.480
<v Speaker 1>It's just two bars at the beginning, UM with ascendant

0:04:57.120 --> 0:05:03.120
<v Speaker 1>melodique movement. That's it is really unexpected and that I loved.

0:05:03.440 --> 0:05:07.360
<v Speaker 1>I was at the first first time. So he asked

0:05:07.360 --> 0:05:10.800
<v Speaker 1>the system to generate new ones based on those new parameters,

0:05:11.080 --> 0:05:15.480
<v Speaker 1>and then I reiterate with the machine like you know,

0:05:15.560 --> 0:05:20.080
<v Speaker 1>if if it were a workmate something like that. It's

0:05:20.120 --> 0:05:24.039
<v Speaker 1>kind of like, um, two artists collaborating on a song,

0:05:24.160 --> 0:05:27.320
<v Speaker 1>but it's just that one of them is a machine exactly.

0:05:27.720 --> 0:05:31.240
<v Speaker 1>And so that ultimately led to this full song that

0:05:31.320 --> 0:05:39.640
<v Speaker 1>he released under his artist name Skeige. It's it's ding

0:05:40.080 --> 0:06:02.640
<v Speaker 1>those doings are those use Yeah, I definitely recognized the

0:06:02.680 --> 0:06:06.479
<v Speaker 1>melody from earlier, but it doesn't sound as jazzy. It

0:06:06.520 --> 0:06:10.600
<v Speaker 1>sounds some a little old timey and familiar. At the end,

0:06:10.680 --> 0:06:16.200
<v Speaker 1>I had really, uh, very interesting melody. I was able

0:06:16.240 --> 0:06:20.680
<v Speaker 1>to say, Okay, I'm proud of this melody and I

0:06:20.680 --> 0:06:25.240
<v Speaker 1>think it's new. I think it's interesting because I couldn't

0:06:25.279 --> 0:06:31.800
<v Speaker 1>have made it by myself. So the collaboration with AI

0:06:32.160 --> 0:06:37.360
<v Speaker 1>is really interesting because it gives me something new. So

0:06:37.880 --> 0:06:41.280
<v Speaker 1>for this song, the AI was really just the initial inspiration,

0:06:42.040 --> 0:06:45.760
<v Speaker 1>but Benuah made a ton of creative decisions to bring

0:06:45.760 --> 0:06:48.000
<v Speaker 1>this song to the finish line. So in that sense,

0:06:48.040 --> 0:06:51.240
<v Speaker 1>maybe you can call this AI light right. And so

0:06:51.320 --> 0:06:56.520
<v Speaker 1>Bena released this song back in and since then, Francois

0:06:56.560 --> 0:07:00.280
<v Speaker 1>Pachet from Sony moved over to Spotify, and Ben has

0:07:00.320 --> 0:07:04.680
<v Speaker 1>been helping francoise new work at Spotify and they're they're

0:07:04.680 --> 0:07:09.120
<v Speaker 1>building these tools from scratch with researchers for various tasks

0:07:09.200 --> 0:07:12.440
<v Speaker 1>that are involved in creating a song. And one of

0:07:12.480 --> 0:07:16.520
<v Speaker 1>those tools that they've developed, they've incorporated into Ben's new

0:07:16.560 --> 0:07:20.280
<v Speaker 1>album called American Folk Songs. So been one and I

0:07:20.360 --> 0:07:23.120
<v Speaker 1>talked about one song called black is a Color. The

0:07:23.160 --> 0:07:26.160
<v Speaker 1>melody is based on this famous folk song called black

0:07:26.240 --> 0:07:28.560
<v Speaker 1>is the Color of My True Loves hair. Um. It's

0:07:28.600 --> 0:07:32.240
<v Speaker 1>been covered by people like Pete Seeger Nina Simone. I'm

0:07:32.240 --> 0:07:36.160
<v Speaker 1>sure you've heard it before. I actually don't think I have.

0:07:36.640 --> 0:07:39.920
<v Speaker 1>UM is that bad? I grew up into dead well,

0:07:40.040 --> 0:07:43.600
<v Speaker 1>I think most of our listeners have. So Ben Wass

0:07:43.600 --> 0:07:47.360
<v Speaker 1>song is based on this acapella version sung by Pete Seeger.

0:07:47.720 --> 0:07:55.080
<v Speaker 1>Black Black, Black, Color of My True loves hand and

0:07:55.120 --> 0:07:58.200
<v Speaker 1>he had Spotify's AI program. I'll add all these harmonies

0:07:58.200 --> 0:08:00.360
<v Speaker 1>in the background to make it sound so much a chair.

0:08:00.720 --> 0:08:07.640
<v Speaker 1>Black black, black is the color of my true love's head.

0:08:09.640 --> 0:08:17.280
<v Speaker 1>Her face is something wondrous fair. But yourlest eyes and

0:08:17.520 --> 0:08:29.720
<v Speaker 1>the daintiest hands. I love the ground she STIs. I love.

0:08:30.720 --> 0:08:32.800
<v Speaker 1>I thought that it could be very black in Sparring,

0:08:32.880 --> 0:08:36.440
<v Speaker 1>to have something very sophisticated with a simple melody like

0:08:36.720 --> 0:08:39.679
<v Speaker 1>black is the color, and I got a really unnam

0:08:39.760 --> 0:08:43.080
<v Speaker 1>is amazing result. I was blown away by this result.

0:08:44.559 --> 0:08:51.840
<v Speaker 1>Black black, black is the color of my true loves

0:08:51.960 --> 0:09:04.400
<v Speaker 1>her her face is something wondrous fair. I think, no, no,

0:09:04.400 --> 0:09:08.600
<v Speaker 1>no human could have composed this string arrangement because it's

0:09:09.040 --> 0:09:12.520
<v Speaker 1>too weird. There are two strange things in it too,

0:09:12.600 --> 0:09:20.440
<v Speaker 1>strange chord changes and internal movements. But it's still really,

0:09:20.480 --> 0:09:25.920
<v Speaker 1>really beautiful. And I think that this kind of result

0:09:26.679 --> 0:09:32.040
<v Speaker 1>is very encouraging for for musicians like me and after me,

0:09:32.720 --> 0:09:38.280
<v Speaker 1>musicians did something new, something that they find in Sparring

0:09:38.400 --> 0:09:54.920
<v Speaker 1>for their compositions. You know, it's really different from the

0:09:55.040 --> 0:09:59.240
<v Speaker 1>music i'man're still listening to, uh and definitely a little weird,

0:09:59.360 --> 0:10:02.680
<v Speaker 1>but I think I like it. It's beautiful. Yeah, it

0:10:02.760 --> 0:10:05.080
<v Speaker 1>sounds I don't know, both old and knew at the

0:10:05.120 --> 0:10:07.680
<v Speaker 1>same time. But what's interesting to me is how he's

0:10:07.720 --> 0:10:11.760
<v Speaker 1>really revamping this traditional folk music, and I can imagine

0:10:12.160 --> 0:10:14.920
<v Speaker 1>it might sound strange for some American audiences who grew

0:10:15.000 --> 0:10:17.880
<v Speaker 1>up listening to this type of music and then have

0:10:18.040 --> 0:10:22.480
<v Speaker 1>it repackaged in a totally different way. So the original

0:10:22.520 --> 0:10:26.520
<v Speaker 1>melody and the lyrics were obviously written by a human

0:10:27.120 --> 0:10:30.680
<v Speaker 1>a long time ago, but the arrangement we hear in

0:10:30.720 --> 0:10:33.439
<v Speaker 1>the background was all composed by a machine. Is that right?

0:10:34.000 --> 0:10:37.040
<v Speaker 1>Mostly so, Ben Watt did make a few choices and

0:10:37.240 --> 0:10:40.400
<v Speaker 1>editions along the way. He used Pete Seeger's voice to

0:10:40.440 --> 0:10:43.719
<v Speaker 1>create an au ac choir, for instance. He also composed

0:10:43.720 --> 0:10:47.959
<v Speaker 1>and recorded some chord sequences. One was in the style

0:10:48.160 --> 0:10:52.560
<v Speaker 1>of like an epic Game of Thrones type sound. Another

0:10:52.600 --> 0:10:55.560
<v Speaker 1>one was in a Bossa Nova style, and he fed

0:10:55.600 --> 0:10:58.320
<v Speaker 1>that into the machine, which in turn generated a whole

0:10:58.360 --> 0:11:02.200
<v Speaker 1>new string arrangement for the background on sound and for

0:11:02.240 --> 0:11:05.320
<v Speaker 1>the final recording. He also chose the string quartet and

0:11:05.360 --> 0:11:08.800
<v Speaker 1>the director to play that new composition. Okay, so still

0:11:08.840 --> 0:11:12.440
<v Speaker 1>a lot of human intervention there, Yeah, exactly. Uh, And

0:11:12.480 --> 0:11:15.240
<v Speaker 1>this is a tool that Spotify is making available to everyone,

0:11:15.840 --> 0:11:18.400
<v Speaker 1>so not quite yet. They eventually want to release the

0:11:18.480 --> 0:11:21.320
<v Speaker 1>tool on open source on the Internet, possibly by the

0:11:21.400 --> 0:11:23.800
<v Speaker 1>end of next year, but it could take longer than that.

0:11:27.200 --> 0:11:29.560
<v Speaker 1>And so the next composer and entrepreneur I want to

0:11:29.600 --> 0:11:33.040
<v Speaker 1>introduce you to is this guy called Pierre Bahl and

0:11:33.120 --> 0:11:37.360
<v Speaker 1>he's based in Luxembourg. He studied computer science. He's a musician,

0:11:37.640 --> 0:11:40.160
<v Speaker 1>his father is a film and music producer, his mother

0:11:40.240 --> 0:11:42.680
<v Speaker 1>is a singer, and he and his brother started this

0:11:42.720 --> 0:11:46.040
<v Speaker 1>company called Ava. This is the startup that made the

0:11:46.120 --> 0:11:48.760
<v Speaker 1>Divor Jack inspired symphony that we heard at the top

0:11:48.760 --> 0:11:51.199
<v Speaker 1>of the show right, and he and his brother were

0:11:51.240 --> 0:11:54.880
<v Speaker 1>inspired by how important music is in film, but how

0:11:54.920 --> 0:11:58.160
<v Speaker 1>long that process can take in terms of creating a soundtrack.

0:11:58.679 --> 0:12:00.680
<v Speaker 1>So they wanted to see if they could train AI

0:12:00.960 --> 0:12:04.000
<v Speaker 1>and see if it could basically help a composer create

0:12:04.040 --> 0:12:07.200
<v Speaker 1>that type of music, not just for films, but also

0:12:07.280 --> 0:12:12.120
<v Speaker 1>for commercials, promotional videos, video games, and also just generally

0:12:12.320 --> 0:12:16.120
<v Speaker 1>assisting composers in their work. So purefed the machine with

0:12:16.360 --> 0:12:20.920
<v Speaker 1>thirty thou scores of history's greatest composers from Bach, Bitovin,

0:12:21.040 --> 0:12:25.719
<v Speaker 1>and Mozart. So basically it all starts by teaching and

0:12:25.840 --> 0:12:30.680
<v Speaker 1>algorithm to to learn the patterns and music. You know

0:12:30.760 --> 0:12:34.160
<v Speaker 1>that there's this common knowledge that music is sort of

0:12:34.200 --> 0:12:37.520
<v Speaker 1>emotional and it's it's the opposite of math. But actually,

0:12:37.559 --> 0:12:40.679
<v Speaker 1>if you look at music very carefully, there's a lot

0:12:40.720 --> 0:12:44.600
<v Speaker 1>of patterns in it. Um So AVA understands all these

0:12:44.640 --> 0:12:47.680
<v Speaker 1>patterns which are very very mathematical at the core, and

0:12:47.760 --> 0:12:50.880
<v Speaker 1>EVA uses that to generate different kinds of music depending

0:12:50.880 --> 0:12:54.000
<v Speaker 1>on what you're looking for. So let me play you

0:12:54.160 --> 0:12:57.440
<v Speaker 1>two songs in a completely different style. The first one

0:12:57.600 --> 0:13:21.320
<v Speaker 1>is this classical song called I Am a h This

0:13:21.440 --> 0:13:26.760
<v Speaker 1>makes me think of a winter ballet. Does that make sense? Absolutely? Yeah,

0:13:27.040 --> 0:13:32.160
<v Speaker 1>I totally think of forest blanketed and snow and little

0:13:32.200 --> 0:13:35.520
<v Speaker 1>bunny rabbits popping by. I see that too. Let me

0:13:35.559 --> 0:13:39.480
<v Speaker 1>play a totally different song. This is a pop song

0:13:39.559 --> 0:13:53.720
<v Speaker 1>called Guiding Light. So background music is a big industry,

0:13:54.080 --> 0:13:58.200
<v Speaker 1>and this just underscores the strength of Ava's business model.

0:13:58.440 --> 0:14:00.319
<v Speaker 1>I mean, for the most part, if you're come penny

0:14:00.520 --> 0:14:03.440
<v Speaker 1>or like a podcast, you can pay a composer to

0:14:03.480 --> 0:14:06.000
<v Speaker 1>make custom music, or you can use these catalogs with

0:14:06.120 --> 0:14:08.600
<v Speaker 1>songs that you're licensed to use. Right, That's what we

0:14:08.679 --> 0:14:13.240
<v Speaker 1>do for this show. We use these big catalogs of

0:14:13.360 --> 0:14:16.920
<v Speaker 1>songs UM that we're allowed to use, but it's really

0:14:16.960 --> 0:14:20.320
<v Speaker 1>hard because you know, these are already existing songs. A

0:14:20.320 --> 0:14:22.760
<v Speaker 1>lot of other shows use them, and I don't know,

0:14:22.800 --> 0:14:25.240
<v Speaker 1>I know, I drive our producers crazy because I'm really

0:14:25.280 --> 0:14:28.360
<v Speaker 1>picky about which tracks they use for different sections of

0:14:28.360 --> 0:14:31.440
<v Speaker 1>our show. Yeah, yeah, I mean, so that's the big

0:14:31.480 --> 0:14:35.080
<v Speaker 1>opportunity for EVA. You could have a custom made piece

0:14:35.120 --> 0:14:37.520
<v Speaker 1>of music, but because you're having a computer do it,

0:14:37.520 --> 0:14:41.960
<v Speaker 1>it's probably cheaper and faster than having a human composer

0:14:42.000 --> 0:14:45.280
<v Speaker 1>do it. How long does it take to create like

0:14:45.320 --> 0:14:47.800
<v Speaker 1>a new song, you know, like three minutes, three and

0:14:47.840 --> 0:14:50.560
<v Speaker 1>a half minutes, So it depends. It depends with the

0:14:50.560 --> 0:14:52.960
<v Speaker 1>algorithms that we use, because we have different algorithms that

0:14:53.040 --> 0:14:56.320
<v Speaker 1>we've used of the basket full of years. The very

0:14:56.360 --> 0:14:59.720
<v Speaker 1>first ones that we had it took usually forty eight

0:14:59.720 --> 0:15:02.320
<v Speaker 1>to say many two hours because we had to retrain

0:15:03.160 --> 0:15:09.480
<v Speaker 1>av the influences that we wanted her to specifically emanate UM.

0:15:09.560 --> 0:15:12.800
<v Speaker 1>But more recently it takes about like forty five seconds

0:15:12.840 --> 0:15:15.440
<v Speaker 1>to a minute to create a piece of music from

0:15:15.640 --> 0:15:20.160
<v Speaker 1>three minutes long. Now, so you know how earlier the

0:15:20.240 --> 0:15:24.320
<v Speaker 1>first example from ben Wa was just this starting point

0:15:24.760 --> 0:15:27.880
<v Speaker 1>and there was still like a ton of human intervention

0:15:28.000 --> 0:15:32.440
<v Speaker 1>that was involved. How complete is the music that Ava's

0:15:32.480 --> 0:15:36.720
<v Speaker 1>software is generating, well, Pierce, is it kind of depends

0:15:36.760 --> 0:15:39.680
<v Speaker 1>on who's playing around with the tool, Like someone who's

0:15:39.880 --> 0:15:43.200
<v Speaker 1>highly musically trained might make more tweaks and changes to it,

0:15:43.280 --> 0:15:46.920
<v Speaker 1>whereas someone who isn't might leave it as is, though

0:15:46.920 --> 0:15:48.960
<v Speaker 1>the quality might not be as good as a result.

0:15:49.560 --> 0:15:53.160
<v Speaker 1>So in the songs that we just listened to earlier,

0:15:54.360 --> 0:15:58.520
<v Speaker 1>did human composers tweak those songs? So nope, that was

0:15:58.680 --> 0:16:04.160
<v Speaker 1>composed entirely by Eva, but humans did assign the different

0:16:04.200 --> 0:16:26.600
<v Speaker 1>parts of music two different instruments. We'll be right back, Okay,

0:16:26.680 --> 0:16:29.640
<v Speaker 1>So before the break, we met two musicians who are

0:16:29.720 --> 0:16:33.520
<v Speaker 1>using AI to make music in really different ways. Yeah,

0:16:33.560 --> 0:16:37.680
<v Speaker 1>so Bena uses AI kind of as a collaborator, Pierre

0:16:37.680 --> 0:16:42.440
<v Speaker 1>Berrow uses AI specifically for background music. But the last

0:16:42.440 --> 0:16:44.880
<v Speaker 1>guy I want to introduce you to takes it even further.

0:16:45.240 --> 0:16:49.480
<v Speaker 1>His name's ash Kusha. Yeah, thanks so much for for

0:16:49.680 --> 0:16:54.000
<v Speaker 1>having us sober. We're really and he's an electronic musician.

0:16:54.080 --> 0:16:57.960
<v Speaker 1>He's Iranian born and based in London. The question for

0:16:58.080 --> 0:17:02.680
<v Speaker 1>me always was how I can I can make very intriguing,

0:17:03.160 --> 0:17:07.760
<v Speaker 1>complete and complex piece of music only using a computer.

0:17:08.359 --> 0:17:11.919
<v Speaker 1>And also what if the part of the composition that

0:17:11.960 --> 0:17:15.800
<v Speaker 1>I'm thinking of can be partly made by the computer.

0:17:16.480 --> 0:17:20.160
<v Speaker 1>So that question took me twelve years to go through

0:17:20.200 --> 0:17:23.520
<v Speaker 1>all the different parts, from instruments, replicating sound of violence,

0:17:24.080 --> 0:17:29.320
<v Speaker 1>field recording, synthesizing voice, and finally to finding a way

0:17:29.359 --> 0:17:35.120
<v Speaker 1>to generate lyrics. Um, that is what initiated oxyen. It

0:17:35.240 --> 0:17:41.919
<v Speaker 1>was trying to tackle what is a non human creative engine.

0:17:42.200 --> 0:17:45.800
<v Speaker 1>And a few years ago Ash started building virtual entertainers

0:17:45.840 --> 0:17:47.960
<v Speaker 1>for video games. So you know how they have these

0:17:47.960 --> 0:17:52.399
<v Speaker 1>avatars and video games, is that like Mario and Mario

0:17:52.520 --> 0:17:57.680
<v Speaker 1>kart am I adding itself is a non gamer. Well,

0:17:57.680 --> 0:18:01.480
<v Speaker 1>so Ash made a virtual avatar that makes music and

0:18:01.720 --> 0:18:04.639
<v Speaker 1>he called her Yonah. So this is like a whole

0:18:05.359 --> 0:18:09.359
<v Speaker 1>virtual character that writes her own music. Yeah, so you

0:18:09.400 --> 0:18:13.320
<v Speaker 1>can kind of think of it as infusing an AI

0:18:13.400 --> 0:18:17.080
<v Speaker 1>with a personality too. So in this case, Ash trained

0:18:17.160 --> 0:18:21.120
<v Speaker 1>Yonah's AI on Margaret Atwood's novels and also on articles

0:18:21.160 --> 0:18:25.840
<v Speaker 1>about teenage life. So Ash's company ox Human created Yonah,

0:18:26.359 --> 0:18:30.840
<v Speaker 1>and she's an angsty teen in a dystopian world. That's

0:18:30.880 --> 0:18:33.480
<v Speaker 1>a really good way of putting it. So, yeah, let

0:18:33.480 --> 0:18:36.359
<v Speaker 1>me play you a song that Ash just released in September,

0:18:36.480 --> 0:18:41.720
<v Speaker 1>written and sung by Yona. I never felt alone. You

0:18:41.880 --> 0:18:47.359
<v Speaker 1>never said a word. I fell from my prone. You

0:18:47.440 --> 0:18:53.280
<v Speaker 1>didn't want me there, you didn't want me near, You

0:18:53.320 --> 0:19:04.440
<v Speaker 1>didn't want me there. I never heard you say it's

0:19:04.480 --> 0:19:18.960
<v Speaker 1>every life I live. I it's definitely very weird, very dystopian.

0:19:20.440 --> 0:19:21.919
<v Speaker 1>I feel like she's going to come and kill me

0:19:21.960 --> 0:19:25.080
<v Speaker 1>in my sleep. Yeah, I mean it's definitely a little eerie.

0:19:25.240 --> 0:19:27.159
<v Speaker 1>And some of the comments on the video are kind

0:19:27.200 --> 0:19:30.639
<v Speaker 1>of funny, like one user says this is creepy, and

0:19:30.720 --> 0:19:32.800
<v Speaker 1>other one says, don't listen to this when you're high,

0:19:33.680 --> 0:19:37.360
<v Speaker 1>and another guy says, don't mess with organic music. Um,

0:19:37.480 --> 0:19:40.080
<v Speaker 1>And I think part of what makes it extra creepy

0:19:40.200 --> 0:19:42.560
<v Speaker 1>is that you can watch Yona sing this song on

0:19:42.600 --> 0:19:46.600
<v Speaker 1>YouTube and she looks half human half robot. If you cry,

0:19:46.680 --> 0:19:52.400
<v Speaker 1>would every smile, I'll make you stare? I get where

0:19:52.400 --> 0:19:56.280
<v Speaker 1>do you were? Were they all happened? You know? So

0:19:56.640 --> 0:19:59.960
<v Speaker 1>this definitely really creeps me out. But now that I'm

0:20:00.080 --> 0:20:04.520
<v Speaker 1>thinking about it, maybe that's the point that it's disturbing yeah.

0:20:04.520 --> 0:20:08.040
<v Speaker 1>So Ash's goal with Jonah and his other characters he

0:20:08.080 --> 0:20:12.280
<v Speaker 1>has other AI based characters is to really have um

0:20:12.560 --> 0:20:15.280
<v Speaker 1>have them in vogue emotion and the people listening to it.

0:20:15.880 --> 0:20:19.800
<v Speaker 1>So in her music you get this sense that she's sad,

0:20:20.520 --> 0:20:24.120
<v Speaker 1>and Ash says it's easier to get human reaction through

0:20:24.160 --> 0:20:28.040
<v Speaker 1>sad and romantic music. Um, we want to make something

0:20:28.080 --> 0:20:30.320
<v Speaker 1>that when you listen to Yon know you you believe

0:20:30.400 --> 0:20:35.840
<v Speaker 1>that there's something being said. So the nature of of

0:20:36.480 --> 0:20:38.439
<v Speaker 1>the scales that we use, and and the and the

0:20:38.480 --> 0:20:40.560
<v Speaker 1>type of music that we use, it's very personal, it's

0:20:40.640 --> 0:20:43.960
<v Speaker 1>very deep, it's very um yeah, it's in the sense

0:20:44.000 --> 0:20:47.000
<v Speaker 1>is romantic. I think the other thing he says is

0:20:47.040 --> 0:20:50.160
<v Speaker 1>that he's not trying to replace traditional music in any way,

0:20:50.200 --> 0:20:52.800
<v Speaker 1>but to create a whole new form of sound and

0:20:52.800 --> 0:20:55.440
<v Speaker 1>a whole new genre. We we could try and make

0:20:55.480 --> 0:20:59.720
<v Speaker 1>the best rock track, but I think there are many

0:20:59.760 --> 0:21:03.240
<v Speaker 1>many good musicians that do that. Creating genres takes time,

0:21:03.320 --> 0:21:06.760
<v Speaker 1>and it's just a subculture and it's seen pretty much,

0:21:07.000 --> 0:21:08.800
<v Speaker 1>so it takes time, so we want to give it time.

0:21:09.080 --> 0:21:12.320
<v Speaker 1>At the beginning of soundings sometimes wonky and it's a

0:21:12.320 --> 0:21:16.000
<v Speaker 1>bit weird, but that's what we're looking for. And so

0:21:16.040 --> 0:21:20.120
<v Speaker 1>with that also goes a different form of consumption. So

0:21:22.000 --> 0:21:24.800
<v Speaker 1>this type of music might be consumed through games or

0:21:24.880 --> 0:21:28.159
<v Speaker 1>virtual reality or different types of concerts, and there's the

0:21:28.320 --> 0:21:32.000
<v Speaker 1>market for that. Yeah. I mean he's already ashes already

0:21:32.000 --> 0:21:37.200
<v Speaker 1>monetizing his virtual entertainers. I mean Jonah performed at events

0:21:37.240 --> 0:21:41.439
<v Speaker 1>and shows and at a poetry festival recently. Wow. So

0:21:41.480 --> 0:21:45.000
<v Speaker 1>it's it's kind of like um hatsen Amiku and Japan.

0:21:45.320 --> 0:21:47.840
<v Speaker 1>Do you know about her? I have no idea who

0:21:47.920 --> 0:21:50.679
<v Speaker 1>that is. So she's just like they call her a

0:21:50.800 --> 0:21:54.919
<v Speaker 1>virtual pop idol um but it's it's not like they

0:21:55.000 --> 0:21:57.919
<v Speaker 1>have like tons of people who go to her concerts

0:21:58.040 --> 0:22:00.560
<v Speaker 1>or her or it or whatever you would call it.

0:22:09.760 --> 0:22:22.560
<v Speaker 1>M Yeah, and she gets projected onto a screen and

0:22:22.920 --> 0:22:26.359
<v Speaker 1>you know, they use the software for her um to

0:22:26.960 --> 0:22:30.000
<v Speaker 1>to sing, if you can call it that. That is

0:22:30.080 --> 0:22:34.320
<v Speaker 1>the idea of the modern, the contemporary of pop culture,

0:22:34.359 --> 0:22:38.040
<v Speaker 1>which is alter egos and creating almost cartoon characters. So

0:22:38.080 --> 0:22:40.960
<v Speaker 1>we we are kind of adopting all of these things

0:22:41.000 --> 0:22:44.720
<v Speaker 1>into a new form of expression. This is why we're

0:22:44.760 --> 0:22:48.200
<v Speaker 1>not apologetic about, you know, not being digital or being

0:22:48.240 --> 0:22:51.200
<v Speaker 1>a bit weird and not complete or perfect. That's how

0:22:51.240 --> 0:22:55.280
<v Speaker 1>she is. And I think, um, the gaming generation, the

0:22:55.400 --> 0:22:59.040
<v Speaker 1>video game generation and Juris and Alpha is going to

0:22:59.080 --> 0:23:10.560
<v Speaker 1>be more except thing. So now we started this episode

0:23:10.600 --> 0:23:15.480
<v Speaker 1>with this idea that creativity is the last frontier of AI,

0:23:16.000 --> 0:23:18.360
<v Speaker 1>and of course that begs the question, if AI can

0:23:18.359 --> 0:23:21.919
<v Speaker 1>now make music, is there anything left for us humans

0:23:21.920 --> 0:23:25.879
<v Speaker 1>to do? And you know, having listened to your conversations

0:23:26.040 --> 0:23:29.199
<v Speaker 1>with Benoir and Pierre and Ash and having listened to

0:23:29.240 --> 0:23:33.400
<v Speaker 1>their AI generated music, I think it's all really interesting

0:23:33.440 --> 0:23:36.600
<v Speaker 1>and there were definitely sections of their songs that I

0:23:36.640 --> 0:23:39.359
<v Speaker 1>really liked. But the question we posed at the start

0:23:39.520 --> 0:23:42.560
<v Speaker 1>feels still very premature to me. It feels like we're

0:23:42.600 --> 0:23:47.760
<v Speaker 1>still very very far from AI replacing human composers. And

0:23:47.840 --> 0:23:51.480
<v Speaker 1>I personally still prefer all of the artists I listened

0:23:51.520 --> 0:23:54.480
<v Speaker 1>to who are writing their own stuff instead of having

0:23:54.480 --> 0:23:57.680
<v Speaker 1>computers write their stuff. You know what. It reminds me

0:23:58.080 --> 0:24:01.359
<v Speaker 1>of the conversation that I had with a Barrow from Ava,

0:24:01.480 --> 0:24:05.800
<v Speaker 1>and he said that people were really upset when synthesizers

0:24:05.800 --> 0:24:08.840
<v Speaker 1>were first introduced thirty years ago, and now it's totally normal.

0:24:09.359 --> 0:24:11.600
<v Speaker 1>So I guess I can imagine a day when I

0:24:11.640 --> 0:24:14.040
<v Speaker 1>was just going to be another tool for musicians and

0:24:14.160 --> 0:24:26.160
<v Speaker 1>people won't even bat an eye. Natalia drags Jack. Thanks

0:24:26.160 --> 0:24:28.879
<v Speaker 1>for coming on the show today. Thanks for having me Aki.

0:24:33.840 --> 0:24:35.560
<v Speaker 1>So before I let you guys go, I want to

0:24:35.640 --> 0:24:37.840
<v Speaker 1>let you know that this is the last episode of

0:24:37.880 --> 0:24:40.960
<v Speaker 1>this season of Decrypted, and we're going to be publishing

0:24:40.960 --> 0:24:44.679
<v Speaker 1>a special bonus episode in two weeks where this show's

0:24:44.720 --> 0:24:47.240
<v Speaker 1>original co host, Brad Stone and I are going to

0:24:47.320 --> 0:24:50.240
<v Speaker 1>be talking about our favorite episodes that we've ever done

0:24:50.600 --> 0:24:52.680
<v Speaker 1>and update you on the people and the stories we've

0:24:52.680 --> 0:24:55.439
<v Speaker 1>covered over the years. So if there's an episode you

0:24:55.440 --> 0:24:58.679
<v Speaker 1>want to talk about, tweet at Meat at seven or

0:24:58.760 --> 0:25:01.760
<v Speaker 1>email me a I t O one six at Bloomberg

0:25:01.800 --> 0:25:06.639
<v Speaker 1>dot net. Decrypted is hosted by me. Shawn ween is

0:25:06.640 --> 0:25:10.760
<v Speaker 1>our executive producer. Ethan Brooks mixes show today. Nate Linkson

0:25:10.840 --> 0:25:14.439
<v Speaker 1>and Neville Gillette helped with recordings. Francesca Levy is the

0:25:14.480 --> 0:25:17.760
<v Speaker 1>head of Bloomberg Podcasts. We'll see you in two weeks