WEBVTT - Artificial Intelligence vs. The Music Industry 0:00:00.600 --> 0:00:02.800 I think a lot of people think it's just going 0:00:02.880 --> 0:00:04.760 to be great. If we just make these things, they're 0:00:04.800 --> 0:00:06.320 just going to be great, and we're fine. And a 0:00:06.320 --> 0:00:08.840 lot of people think this is definitely a dystopia, and 0:00:08.840 --> 0:00:12.000 we're definitely for it, and I think it can truly 0:00:12.119 --> 0:00:13.280 be a third thing. 0:00:14.680 --> 0:00:18.400 Welcome to Crash Course, a podcast about business, political, and 0:00:18.440 --> 0:00:21.680 social disruption and what we can learn from it. I'm 0:00:21.720 --> 0:00:28.040 Tim O'Brien. Today's Crash Course Artificial Intelligence versus the Music Industry. 0:00:29.160 --> 0:00:31.200 The voice you heard at the open was producer and 0:00:31.280 --> 0:00:36.080 songwriter Extraordinary Grimes speaking to us about the good, the bad, 0:00:36.280 --> 0:00:40.280 and the in between of artificial intelligence and music. The 0:00:40.360 --> 0:00:43.159 AI hype may be starting to die down overall, but 0:00:43.280 --> 0:00:47.120 AI isn't going away. You can literally hear it in music. 0:00:47.760 --> 0:00:51.360 Back in April, a super convincing AI duet between rapper 0:00:51.440 --> 0:00:55.080 Drake and singer The Weekend went viral, causing panic in 0:00:55.120 --> 0:00:58.440 the music industry, and it opened the floodgates for deep 0:00:58.480 --> 0:01:01.800 fakes like Frank Sinatra singh hip hop and a fictional 0:01:01.840 --> 0:01:07.360 Oasis mixtape called Aisis. It's not just novelty memes either. 0:01:08.000 --> 0:01:11.839 AI generated background music is sneaking into your Spotify playlist, 0:01:12.319 --> 0:01:15.920 cracking out chill soundscapes, and helping the Beatles put out 0:01:15.959 --> 0:01:19.960 one last song called Now and Then. So music is 0:01:20.000 --> 0:01:23.160 again on the cusp of huge tech driven change. Twenty 0:01:23.240 --> 0:01:27.600 years after file sharing misfit Nabster made CD collections obsolete. 0:01:28.440 --> 0:01:30.640 To help us delve into what that change might look like, 0:01:31.200 --> 0:01:34.440 we at Bloomberg Opinion set one of our columnist detectives, 0:01:34.760 --> 0:01:38.920 Lionelle Laurent, out to investigate. Lionelle is a writer who 0:01:38.920 --> 0:01:41.640 wears many hats, one of which involves keeping an eye 0:01:41.680 --> 0:01:45.880 on digital innovation and upheaval in the music industry, and 0:01:45.920 --> 0:01:50.400 he has a tale to tell. Hi, Lianel, Hi, sim 0:01:50.440 --> 0:01:52.360 I'll let our listeners know that you've been working on 0:01:52.400 --> 0:01:56.440 this from Paris, So before we get started, why music well. 0:01:56.640 --> 0:01:59.360 Music is a personal passion of mine, and music is 0:01:59.400 --> 0:02:02.120 part of what makes as humans, So seeing AI take 0:02:02.120 --> 0:02:05.000 on more of that is I think a big cultural moment. 0:02:05.520 --> 0:02:07.920 But music as an industry has also served as a 0:02:08.000 --> 0:02:11.160 kind of canarian, the coal mine for wider tech disruption, 0:02:11.440 --> 0:02:14.080 including streaming. It really holds a mirror up to the 0:02:14.080 --> 0:02:16.240 rest of the economy, so I think it's important for 0:02:16.280 --> 0:02:18.120 everyone to follow this well. 0:02:18.240 --> 0:02:19.840 I'm going to hand it over to you to tell 0:02:19.880 --> 0:02:21.800 us what you've learned, and I'll catch up with you 0:02:21.880 --> 0:02:22.560 later in the show. 0:02:23.080 --> 0:02:26.480 Thanks Tim. Now, I'm willing to bet that you probably 0:02:26.480 --> 0:02:30.040 haven't heard of French songwriter bunoits Care. He's got a 0:02:30.080 --> 0:02:33.840 special claim to music fame. He's behind the first ever 0:02:33.919 --> 0:02:38.200 song composed by artificial intelligence. It's called Daddy's Car and 0:02:38.280 --> 0:02:40.040 it was released in twenty sixteen. 0:02:58.520 --> 0:03:04.200 In Daddy's Car Sound. So do we like something new? 0:03:04.760 --> 0:03:05.400 It ms? 0:03:07.320 --> 0:03:10.120 This isn't one hundred percent AI. Bern Noir wrote the 0:03:10.160 --> 0:03:13.040 lyrics and arranged the song, and that's him singing, but 0:03:13.080 --> 0:03:16.160 the rest was composed by an AI model trained entirely 0:03:16.240 --> 0:03:19.960 on songs by the Beatles, using all those Fab four classics. 0:03:20.240 --> 0:03:22.840 The machine came up with its own suggestions and ideas. 0:03:23.320 --> 0:03:25.560 It's like if John, Paul, George and Ringo had a 0:03:25.560 --> 0:03:29.560 fifth robotic friend jamming away and Bernoit playing with the results. 0:03:31.960 --> 0:03:34.600 I played with the tool. I fed it with fifty 0:03:34.760 --> 0:03:37.760 songs of the Beatles, my favorite songs of the Beatles, 0:03:38.040 --> 0:03:40.880 and then I got a result eight bars and then 0:03:40.880 --> 0:03:43.840 it was funny, and I continued. 0:03:48.960 --> 0:03:51.480 So I think we can agree this AI music isn't 0:03:51.560 --> 0:03:54.280 quite at the level of let it be Still. Daddy's 0:03:54.320 --> 0:03:57.080 car was a kind of milestone, and for bu Noir 0:03:57.240 --> 0:04:00.600 it is like improvising or writing a regular song hunched 0:04:00.640 --> 0:04:03.320 over a piano. He's been using tech to make music 0:04:03.320 --> 0:04:06.000 since the days of the Commodore sixty four kids ask 0:04:06.040 --> 0:04:08.440 your parents what that is today. He does it all 0:04:08.440 --> 0:04:11.960 from his studio in Paris, a music geeks paradise, stuffed 0:04:11.960 --> 0:04:16.040 with guitars, real real tape and computers. It all helps 0:04:16.080 --> 0:04:19.360 inspire new ideas, which he puts out under the name Skiggit. 0:04:19.600 --> 0:04:22.880 That's Danish for shadow, inspired by a Haunds Christian Anderson 0:04:22.920 --> 0:04:33.000 tale about a shadow that's alive. 0:04:36.320 --> 0:04:42.760 AI helped me to explore an expected part of my creativity. 0:04:43.640 --> 0:04:48.400 It's like the shadow in the Tale of Anderssan with 0:04:49.080 --> 0:04:53.560 becoming something independent from the main character. 0:05:04.760 --> 0:05:07.800 This summer, Bnoir gave me a magical mystery tour of 0:05:07.880 --> 0:05:11.320 how AI could revolutionize music and why it shouldn't be 0:05:11.360 --> 0:05:15.440 banned or dismissed out of hand. First stop deep fakes 0:05:15.920 --> 0:05:19.359 like Heart on My Sleeve that Drake Weekend duet we 0:05:19.440 --> 0:05:27.480 mentioned earlier. Bernoir doesn't think the imitation game is all 0:05:27.480 --> 0:05:31.080 that interesting musically or fair ethically, but he says it 0:05:31.120 --> 0:05:33.880 shows how AI has improved from the days of Daddy's 0:05:33.880 --> 0:05:37.560 Car to do some things incredibly well, like fooling the 0:05:37.600 --> 0:05:42.440 world into thinking a tambra transfer or vocal transfer is genuine. 0:05:42.720 --> 0:05:45.680 Here's Bernoir whipping up a similar kind of song in 0:05:45.760 --> 0:05:47.279 the style of Eminem. 0:05:47.640 --> 0:05:50.599 I asked Judge Appetty to write a song in the 0:05:50.680 --> 0:05:54.880 style of Eminem. I took this this result from jud 0:05:54.920 --> 0:06:00.279 Jipauty and I sung it and it created this result 0:06:00.440 --> 0:06:03.440 with Eminem's timber and you will hear that there is 0:06:03.480 --> 0:06:07.120 a French excellent. The weight of the world is crushing me. 0:06:07.160 --> 0:06:09.719 Down, trying to keep my head up, but I'm feeling 0:06:09.760 --> 0:06:13.799 as well, the thousand fears they're always lurking in my phrasal. 0:06:13.960 --> 0:06:15.720 Song is kind of hurting. 0:06:15.920 --> 0:06:19.599 Wow, that is pretty convincing, though I should mention we're 0:06:19.600 --> 0:06:21.280 not yet at the level of being able to hit 0:06:21.279 --> 0:06:24.400 a button and get instant deep fake back. This took 0:06:24.440 --> 0:06:28.080 several steps. A voice clone, an AI backing track, lyrics 0:06:28.080 --> 0:06:31.640 by chat GPT and Benoir's vocal cords like the so 0:06:31.720 --> 0:06:34.680 called Mechanical Turk, a fake chess playing robot that was 0:06:34.720 --> 0:06:36.839 all the rage and the seventeen hundreds. There is a 0:06:36.920 --> 0:06:37.800 human hidden in there. 0:06:38.400 --> 0:06:41.600 I got in two hours something that is from nothing 0:06:41.680 --> 0:06:43.040 with the voice of Eminem. 0:06:43.080 --> 0:06:45.960 Deep fakes are just the tip of the AI iceberg, though. 0:06:46.240 --> 0:06:49.479 Bunoir also showed us hybridization, a way to mix his 0:06:49.600 --> 0:06:52.040 voice with that of another singer's, in this case the 0:06:52.120 --> 0:06:55.000 late Chet Baker, and create the kind of weird new 0:06:55.080 --> 0:06:56.320 voice in between. 0:06:58.000 --> 0:07:01.160 Seevu chutt. 0:07:02.800 --> 0:07:03.560 That plan. 0:07:05.040 --> 0:07:09.840 Chutts too. 0:07:14.440 --> 0:07:17.040 The next step will be performers able to use these 0:07:17.160 --> 0:07:20.760 kinds of augmented voices on stage, manipulating not just the 0:07:20.800 --> 0:07:23.360 tone of a sinner's voice, but it's feel, it's groove, 0:07:23.720 --> 0:07:26.280 and eventually also have a machine able to do more, 0:07:26.400 --> 0:07:29.760 maybe all of the composing process. I asked Ben, why 0:07:29.880 --> 0:07:34.920 what it's like as a human dealing with the machine. 0:07:36.720 --> 0:07:39.760 The risk is to lose yourself and get bored by 0:07:39.840 --> 0:07:43.400 all the reasons that you get. Because AI is never tied, 0:07:44.600 --> 0:07:48.520 you can always generate, and I must say that often 0:07:48.600 --> 0:07:51.280 lead the results are not so good. Most of the 0:07:51.320 --> 0:07:54.160 results are like average. It takes time to get the 0:07:54.200 --> 0:07:57.640 result that you were expected, or the result that will 0:07:57.680 --> 0:07:59.880 inspire you to finish the song. 0:08:00.600 --> 0:08:03.119 What's your vision of the future, Then at which point 0:08:03.160 --> 0:08:06.360 do you become obsolete or unnecessary? 0:08:07.080 --> 0:08:10.680 If I can imagine a tool that will make and 0:08:10.800 --> 0:08:14.080 generate what you describe more like me Johne for the 0:08:14.160 --> 0:08:18.800 image for example, or Wali, I think orderly, I mean 0:08:18.920 --> 0:08:21.480 I think it will be great, and maybe it will 0:08:21.520 --> 0:08:25.440 replace some musician or it will be game changer. 0:08:25.920 --> 0:08:28.200 The ability to type in a text prompt and get 0:08:28.200 --> 0:08:32.600 a full song back sounds wild and also pretty dangerous 0:08:32.600 --> 0:08:35.280 when you consider those replaced musicians. But let's face it, 0:08:35.320 --> 0:08:37.680 a sense of danger is not going to slow down 0:08:37.720 --> 0:08:40.719 this tech race, and the finish line may be closer 0:08:40.760 --> 0:08:44.120 than we think. Ed Newton Rex is a composer turned 0:08:44.120 --> 0:08:46.840 AI f Sconado. He says this year will be a 0:08:47.000 --> 0:08:51.079 major milestone towards one hundred percent computer written songs, though 0:08:51.120 --> 0:08:53.440 for him, the real value of AI will be in 0:08:53.480 --> 0:09:07.280 collaborating with and inspiring humans. He points to his own 0:09:07.320 --> 0:09:09.640 work I Stand in the Library, which you can hear 0:09:09.679 --> 0:09:20.040 in the background, as an example of that. He wrote 0:09:20.080 --> 0:09:36.080 the music, but he got AI to write the lyrics. 0:09:43.600 --> 0:09:45.520 This isn't just about kind of creating a three minute 0:09:45.559 --> 0:09:48.400 song having AI do that. Once you have an artificial 0:09:48.480 --> 0:09:49.640 entity that's essentially musical. 0:09:50.040 --> 0:09:50.920 There is so much you can do. 0:09:50.960 --> 0:09:53.200 You can ask it for ideas, you can ask it 0:09:53.200 --> 0:09:56.240 for inspiration, you can have interplay with it, you can 0:09:56.280 --> 0:09:59.319 get it to fill in extra parts in the music 0:09:59.360 --> 0:10:01.280 you're making. You can kind of do anything with it 0:10:01.320 --> 0:10:03.320 that you could do with a human musician working with you. 0:10:03.559 --> 0:10:05.120 And I think that's where it gets really exciting. 0:10:14.360 --> 0:10:18.360 All this enthusiasm for AI has kind of surprised me. Clearly, 0:10:18.520 --> 0:10:22.000 Bonoir and Edge showed that there's real human creativity going 0:10:22.040 --> 0:10:25.040 into AI music. I came into this expectant to find 0:10:25.080 --> 0:10:30.200 tone def tech bros, not real musicians. But I'm also 0:10:30.280 --> 0:10:34.000 feeling uneasy about it. Even setting aside the messy ethics 0:10:34.120 --> 0:10:37.679 raised by AI that's trained on other artists' work, I'm 0:10:37.720 --> 0:10:41.080 starting to get nagging flashbacks of the Disney cartoon Fantasia 0:10:41.240 --> 0:10:44.959 I saw as a kid. Remember the Sources Apprentice, when 0:10:45.000 --> 0:10:48.640 Mickey Mouse starts getting addicted to his productive time saving tool, 0:10:49.120 --> 0:10:52.440 creating an army of broomsticks to do his chores chaos. 0:10:52.440 --> 0:10:56.640 Since hughes as control is lost, and imagining an uncontrolled 0:10:56.679 --> 0:11:00.480 release of AI models and voice clones does not inspire confidence. 0:11:02.280 --> 0:11:04.920 I asked the while what the dangers of this might be. 0:11:05.600 --> 0:11:08.599 I think that Eminem remembers all the songs that he 0:11:08.679 --> 0:11:12.600 has done, so he will easily say no, it's not me. 0:11:14.360 --> 0:11:17.719 So I don't think that this kind of demo, this 0:11:18.120 --> 0:11:22.600 kind of process is really dangerous. But hybotization, as I 0:11:22.720 --> 0:11:26.120 show it begins to be dangerous because you can't control. 0:11:26.800 --> 0:11:29.920 It's hard to control. If I didn't said to you 0:11:30.120 --> 0:11:33.040 like like it's I did it with Chet Baker, you 0:11:33.400 --> 0:11:37.959 wouldn't have recognized Chet Baker, So that becomes a bit 0:11:38.040 --> 0:11:41.560 dangerous here. And it's dangerous also because you can do 0:11:41.600 --> 0:11:42.360 it very easily. 0:11:45.080 --> 0:11:49.400 We have to wonder about unattended consequences like misinformation, misuse, 0:11:49.440 --> 0:11:53.280 and bias too. Last year, a fictional computer generated rapper 0:11:53.360 --> 0:11:56.560 called fn Meeker, complete with green hair and nose rings 0:11:56.559 --> 0:12:00.000 and teeth grills, sparked an online backlash after his lyrics 0:12:00.160 --> 0:12:04.320 used racial slurs and trivialized police brutality. Will the toxic 0:12:04.360 --> 0:12:07.800 content we've seen from aichatbots like chat GPT be any 0:12:07.800 --> 0:12:11.120 more acceptable if it's set to music. This is fundamentally 0:12:11.120 --> 0:12:16.040 about humanity versus technology, says musicians, writes advocate Crispin Hunt 0:12:16.360 --> 0:12:19.240 once a member of nineties British pop band The Long Pigs. 0:12:19.559 --> 0:12:22.640 Paul McCartney singing the Beach Boys is really amusing, Or 0:12:22.760 --> 0:12:25.400 Drake and the Weekend doing a duet on somebody else 0:12:25.559 --> 0:12:28.920 is really amusing. I would be totally freaked out if 0:12:28.960 --> 0:12:32.760 somebody took my voice, and I think we as humans 0:12:32.840 --> 0:12:36.640 need to get in front of that very quickly, before 0:12:37.280 --> 0:12:40.559 before the concrete set before it's set in and before 0:12:41.240 --> 0:12:45.960 mistakes get made, or we just give in to this 0:12:46.160 --> 0:12:50.040 idea that you know, oh it's progress, you can't stop progress. 0:12:50.120 --> 0:12:54.600 But there's a great line by a Polish dissident poet, 0:12:54.760 --> 0:12:57.840 courseeddann islav Lek, which said, is it progress for the 0:12:57.880 --> 0:12:59.080 cannibal eats with a fork? 0:13:01.040 --> 0:13:01.360 Ouch? 0:13:02.040 --> 0:13:04.400 So maybe there is a thin line between the well 0:13:04.480 --> 0:13:08.760 meaning digital composers and the evil fork wielding cannibals. The 0:13:08.880 --> 0:13:12.080 next big question is what is the music industry doing 0:13:12.120 --> 0:13:15.040 about it? Well, the first response from the record labels 0:13:15.040 --> 0:13:18.559 has been straight out of the naps to playbook, fight, Fight, Fight. 0:13:19.400 --> 0:13:21.960 But this isn't the year two thousand anymore. They're good 0:13:21.960 --> 0:13:25.000 at it now. When the NP three landed, record companies 0:13:25.040 --> 0:13:27.400 were like a grandparent trying to understand how to work 0:13:27.400 --> 0:13:30.720 their new smartphone. Today, they've rebuilt themselves around a small 0:13:30.800 --> 0:13:34.240 number of streaming platforms like Spotify. It's a lot easier 0:13:34.280 --> 0:13:37.120 to make digital pirates walk the plank, and there's been 0:13:37.200 --> 0:13:40.440 a lot of plank walking. Hearts on My Sleeve was 0:13:40.520 --> 0:13:44.160 quickly taken down from social media. Unauthorized voice cloning apps 0:13:44.160 --> 0:13:47.760 are being shut down, the Grammys have banned fully AI 0:13:47.960 --> 0:13:51.760 generated music, and Spotify is cracking down on bot activity 0:13:51.800 --> 0:13:55.920 on his platform. Here's Jeffrey Halston, general counsel for Universal 0:13:56.040 --> 0:13:59.200 Music Group, whose artists include Taylor Swift and Ed Sheeran, 0:13:59.600 --> 0:14:02.559 testifying before the US Senate Judiciary Committee in July. 0:14:03.280 --> 0:14:07.920 AI in the service of artists and creativity can be 0:14:07.960 --> 0:14:12.960 a very very good thing. But AI that uses, or 0:14:13.040 --> 0:14:17.400 worse yet, appropriates the work of these artists and creators 0:14:18.200 --> 0:14:22.480 in their creative expression, their name, their image, their likeness, 0:14:22.680 --> 0:14:28.840 their voice without authorization, without consent simply is not. 0:14:29.320 --> 0:14:29.880 A good thing. 0:14:30.680 --> 0:14:33.360 And the music industry has also learned other lessons since 0:14:33.400 --> 0:14:36.120 the napster days. It knows if people really want to 0:14:36.240 --> 0:14:39.480 use something, death by lawsuit is not gonna work, and 0:14:39.600 --> 0:14:42.600 not all AI music is deep fate. Users of AI 0:14:42.640 --> 0:14:47.160 website Boomy have created over seventeen million weird sounding electronic tracks. 0:14:47.160 --> 0:14:49.960 For example, I have trouble getting through just one of them. 0:14:50.000 --> 0:14:52.480 But still maybe one day there'll be a hit in there, 0:14:52.760 --> 0:14:54.760 and not one that can be shut down by a lawyer. 0:14:55.280 --> 0:14:59.120 So here's the next twist. Record labels are done just 0:14:59.200 --> 0:15:02.480 playing defense. They now want to go on the offensive 0:15:02.800 --> 0:15:06.240 by making some AI music of their own that plays 0:15:06.240 --> 0:15:07.080 by their rules. 0:15:07.200 --> 0:15:09.600 Wait, wait, hold it right there, linell. This is shaping 0:15:09.680 --> 0:15:11.480 up to be quite the story, but we need to 0:15:11.480 --> 0:15:13.640 take a quick break and I'll see you on the 0:15:13.640 --> 0:15:14.120 other side. 0:15:14.480 --> 0:15:15.120 Sounds good to me. 0:15:20.880 --> 0:15:24.160 We're back with Lionell Lurn talking about AI and music. 0:15:25.240 --> 0:15:28.160 So Lionelle, let's hear more about how record labels are 0:15:28.160 --> 0:15:30.320 fighting back by co opting AI. 0:15:31.360 --> 0:15:34.680 Sure, it's a story that begins with a young entrepreneur 0:15:34.800 --> 0:15:38.560 based in Berlin called Oleg Stovitski, the AI starts up. 0:15:38.640 --> 0:15:42.280 He co founded Endel, made the music that you're listening 0:15:42.320 --> 0:15:52.600 to now. And Endel has just signed a landmark partnership 0:15:52.880 --> 0:15:56.800 with Universal Music group, that really anti deep fake label 0:15:56.840 --> 0:16:00.320 we heard from earlier, to make AI music use their 0:16:00.400 --> 0:16:03.760 huge catogy stars. And this all started with Oleg looking 0:16:03.760 --> 0:16:07.160 for music that would help them concentrate, similar to the 0:16:07.200 --> 0:16:09.680 work of ambient music pioneer Brian Eno. 0:16:10.440 --> 0:16:13.160 Every time I'm trying to concentrate, I listened to Brain, 0:16:13.240 --> 0:16:16.240 you know, and then noticed that a lot of his music, 0:16:16.240 --> 0:16:19.080 and a lot of ambient music in general started kind 0:16:19.120 --> 0:16:22.640 of popping up. When all of these functional music playlists 0:16:22.680 --> 0:16:26.000 and projects like music for Programming appeared. 0:16:26.520 --> 0:16:29.800 That got Oleg thinking, what if every time you hit play, 0:16:30.240 --> 0:16:33.680 the soundscape was different and fit your mood by monitoring 0:16:33.760 --> 0:16:36.880 data such as your real time heart rate or the 0:16:36.880 --> 0:16:39.400 time of day. That would be like having your own 0:16:39.480 --> 0:16:42.040 personal Brian Eno, made by a machine. 0:16:42.640 --> 0:16:44.320 I kind of turned to my co founders and I 0:16:44.360 --> 0:16:46.880 was like, well, I guess we're going to have to 0:16:46.880 --> 0:16:47.680 build an AI. 0:16:48.360 --> 0:16:51.400 This AI would be trained on a very specific set 0:16:51.400 --> 0:16:55.120 of musical building blocks, or stems. Stems can be thought 0:16:55.120 --> 0:16:58.360 of as these separate instruments or components that make up 0:16:58.360 --> 0:17:04.879 a song. Think of an individual drum beet, a keyboard line, 0:17:07.400 --> 0:17:08.320 a vocal melody. 0:17:08.840 --> 0:17:12.720 I've been to the top of Mount Everest, I've. 0:17:12.560 --> 0:17:17.880 Sailed the seven seats, I've shared the stage with all 0:17:17.960 --> 0:17:19.239 the best. 0:17:20.119 --> 0:17:25.199 As three separate stems, Endel's AI will then uniquely reassemble 0:17:25.240 --> 0:17:29.160 the stems into a new soundscape every time. Personally, it's 0:17:29.160 --> 0:17:31.399 not for me, but clearly there is a market for 0:17:31.440 --> 0:17:34.520 functional music. It represents an estimated seven to ten percent 0:17:34.640 --> 0:17:37.879 of the entire streaming market. Endel's next smart move was 0:17:37.880 --> 0:17:41.600 to blend that robotic AI with well known human artists 0:17:41.680 --> 0:17:45.919 like James Blake and Grimes. They submitted the stems and 0:17:46.119 --> 0:17:49.160 Endal did the rest and basic. This all on stems 0:17:49.520 --> 0:17:53.840 rather than existing albums or songs, means Endel gets official 0:17:53.880 --> 0:17:58.640 recognition as co composer of the real time soundscape. That's 0:17:58.680 --> 0:18:02.160 how Oleg says he avoided a fight with the music industry. 0:18:02.680 --> 0:18:06.080 With us, it was always the reaction was always a 0:18:06.160 --> 0:18:12.159 mix of through kind of genuine interest and almost excitement, 0:18:12.359 --> 0:18:15.960 and people were intrigued because we never removed the artists 0:18:16.000 --> 0:18:18.480 from this process. We never said, oh, you know, we 0:18:18.600 --> 0:18:20.600 just get the stamps from James Blake, but then we're 0:18:20.640 --> 0:18:23.919 going to produce like a million records and James is 0:18:23.960 --> 0:18:25.919 not going to be mentioned anymore. 0:18:26.680 --> 0:18:30.480 But the bigger hand'ed go, the more competition became unavoidable 0:18:30.760 --> 0:18:32.800 because money that would normally be going to the record 0:18:32.840 --> 0:18:36.879 labels is being diverted to text us ups making ambient music. 0:18:37.560 --> 0:18:40.080 The economics of streaming today are all about market share. 0:18:40.480 --> 0:18:43.080 Listen to a track on Spotify for more than thirty seconds, 0:18:43.280 --> 0:18:46.440 and it counts as a play, whether it's forty seconds 0:18:46.440 --> 0:19:00.280 of experimental electronic music, a three minute pop hit, five 0:19:00.320 --> 0:19:08.160 minutes of whale song. The more people click on ambient 0:19:08.240 --> 0:19:12.480 music rather than say, drake, the more it'll impact the 0:19:12.480 --> 0:19:16.440 bottom line. According to Golden Sachs, the three major labels 0:19:16.520 --> 0:19:20.199 market share has dropped from around eighty five percent to 0:19:20.280 --> 0:19:24.520 seventy one percent between twenty seventeen and twenty twenty two, 0:19:24.600 --> 0:19:27.960 which is why when the announcement of that landmark universal 0:19:28.080 --> 0:19:32.320 Endel deal for more record label approved AI music landed 0:19:32.359 --> 0:19:35.080 in May, it looks a lot like can't beat them, 0:19:35.560 --> 0:19:36.439 join them? 0:19:36.680 --> 0:19:39.000 You know they can. I just keep ordering takedowns. That 0:19:39.200 --> 0:19:42.560 is a road to nowhere. They need to have kind 0:19:42.600 --> 0:19:46.879 of an offensive strategy and they need to embrace this 0:19:46.960 --> 0:19:50.280 in some way in form and it's just turned out 0:19:50.280 --> 0:19:52.480 to be that Endel was kind of the only company 0:19:52.600 --> 0:19:55.800 that spent years kind of learning to speak the language, 0:19:55.800 --> 0:19:57.760 and then it's the dance of these labels. 0:19:58.480 --> 0:20:03.040 So what does a record label backed ambient AI do next? 0:20:07.880 --> 0:20:11.920 Ole extreme is to make new AI versions of artists 0:20:12.000 --> 0:20:16.320 existing catalogs. So, for example, Ed Sheeran or Taylor Swift 0:20:16.800 --> 0:20:20.320 might one day offer Endel versions of their albums designed 0:20:20.320 --> 0:20:24.800 for sleep, work or exercise. And imagine if in the future, 0:20:24.880 --> 0:20:29.200 instead of Taylor's version, we get happy, sad or heavy 0:20:29.200 --> 0:20:30.680 metal versions of her albums. 0:20:31.119 --> 0:20:33.919 Essentially expands the universe of the artists. And it's just 0:20:33.960 --> 0:20:39.439 such a beautiful concept of creating a functional soundscape version 0:20:40.160 --> 0:20:41.440 of your favorite artist music. 0:20:41.760 --> 0:20:45.160 It's not just Endel. Other labels are sensing an opportunity. 0:20:45.680 --> 0:20:49.560 Warner Music's CEO recently praised an AI enabled duet between 0:20:49.600 --> 0:20:53.280 Costa Rican musician Pedro Capmani and the generated voice of 0:20:53.280 --> 0:21:01.600 his late father. Oh I asked Gairi, head of music 0:21:01.600 --> 0:21:05.240 group Believe, which owns a DIY music platform called tune Call, 0:21:05.760 --> 0:21:08.840 about this change of heart. Some might say double think 0:21:09.000 --> 0:21:10.240 in the industry. 0:21:10.280 --> 0:21:13.520 Whether it's for Believe or generally for the music industry, 0:21:15.080 --> 0:21:18.560 is an opportunity and not a threat. Why do people 0:21:18.680 --> 0:21:23.600 think AI is a threat generally because they do think 0:21:23.600 --> 0:21:26.520 that the deployment of AI will not be controlled and 0:21:26.600 --> 0:21:30.600 af the opposite view around this, which is that AA 0:21:30.680 --> 0:21:36.040 deployment will be heavily controlled and everyone is approaching AI 0:21:36.640 --> 0:21:39.760 with very a very strong sense of responsibility. 0:21:39.920 --> 0:21:42.960 That control means labels can set their own rules of 0:21:42.960 --> 0:21:46.760 engagement with AI. Deri gives four examples of rules that 0:21:46.800 --> 0:21:50.640 Believe is currently applying on AI. Music uploaded to its platform. 0:21:50.920 --> 0:21:56.240 Consent from the artist, control over the output, compensation, and transparency. 0:21:56.440 --> 0:22:00.119 That's how we're approaching the landscape today, and so we 0:22:00.200 --> 0:22:03.639 are not going to distribute tracks that are one hundred 0:22:03.680 --> 0:22:08.000 percent created by AI unless they meet some of these 0:22:08.040 --> 0:22:08.800 grit areas. 0:22:09.280 --> 0:22:12.400 Believe operates mainly in the mid market of artists rather 0:22:12.400 --> 0:22:15.439 than the superstars, and the hope is that industry friendly 0:22:15.480 --> 0:22:18.760 AI can dig up the next generation's bright young things 0:22:19.200 --> 0:22:21.800 by making it easier to unlock tunes sitting inside their 0:22:21.800 --> 0:22:24.879 heads again or while following the rules. 0:22:25.240 --> 0:22:28.399 We think the number of artists that gets educated for 0:22:28.560 --> 0:22:32.159 technology to become better musician is going to increase, and 0:22:32.560 --> 0:22:36.159 more people creating better music will find an audience. This 0:22:36.320 --> 0:22:40.679 is going to help frontline artists elevate and create better music. 0:22:41.000 --> 0:22:44.359 This optimism around control is also why YouTube has created 0:22:44.400 --> 0:22:47.160 a new set of rules for what it calls responsible 0:22:47.200 --> 0:22:50.719 AI music. But this all seems to be moving very fast. 0:22:50.960 --> 0:22:52.879 There's a lot of pressure on labels to be a 0:22:52.920 --> 0:22:56.520 first mover and sign deals with tech companies. I'll be 0:22:56.600 --> 0:23:00.160 sure that artists consent and all that precious data from 0:23:00.160 --> 0:23:03.639 their life's work isn't being handed over a little too quickly. 0:23:04.440 --> 0:23:08.280 And what if the focus on control starts to stifle creativity? 0:23:08.560 --> 0:23:11.320 If it trips up artists who are using AI but 0:23:11.440 --> 0:23:16.280 fairly without breaching copyright. Remember ed Newton Rex, who he 0:23:16.280 --> 0:23:19.000 heard from earlier in the show. In twenty fifteen, he 0:23:19.080 --> 0:23:22.960 wrapped about this exact thing when promoting his startup jute Deck, 0:23:23.640 --> 0:23:27.080 citing the copyright constraints faced by YouTubers looking for good 0:23:27.080 --> 0:23:27.840 background music. 0:23:28.560 --> 0:23:31.000 Imagine there's a video that you're creating, it's been weeks 0:23:31.000 --> 0:23:33.119 in the making, and your painstaking need take in it 0:23:33.160 --> 0:23:35.200 praying by prayer, have got the whole take looking really 0:23:35.200 --> 0:23:37.960 super great. Start to find the right music. If you've 0:23:37.960 --> 0:23:39.959 got a song you like, you're not allowed to use it. 0:23:40.160 --> 0:23:42.920 Copyright means if you choose it for your video, YouTube 0:23:43.000 --> 0:23:45.920 will just remove it. But this right stuff that's all unique. 0:23:45.920 --> 0:23:47.640 It will do it in any style U. 0:23:47.600 --> 0:23:53.040 See come Back, Fake Drake, all is forgiven? And going 0:23:53.119 --> 0:23:56.520 back to the Sorcerer's Apprentice, what if record label robots 0:23:56.560 --> 0:24:00.199 become so effective at expanding their catalog of stars that 0:24:00.280 --> 0:24:03.040 it makes it harder for the new generation of artists 0:24:03.080 --> 0:24:06.280 to cut through the noise. When does the revolution become stagnation. 0:24:07.280 --> 0:24:09.760 It's time to ask if there's an alternative to either 0:24:09.800 --> 0:24:13.560 the dystopia of no control or the utopia of total control, 0:24:14.280 --> 0:24:16.880 something like a third way that's maybe a little more 0:24:16.880 --> 0:24:20.440 punk rock, putting more power into the hands of the artists. 0:24:20.680 --> 0:24:23.600 And the best person to ask is Grimes, who we 0:24:23.640 --> 0:24:25.560 heard from at the beginning of the show. She is 0:24:25.640 --> 0:24:30.720 one famous artist who is actually embracing AI experimentation while 0:24:30.760 --> 0:24:33.639 also still trying to set limits, sort of like rapper 0:24:33.720 --> 0:24:37.399 Chuck Dee, who during the Napster era said downloadable music 0:24:37.520 --> 0:24:41.080 was a model to embrace. I asked Grimes first off, 0:24:41.119 --> 0:24:44.440 if she generally felt excited or worried about the current 0:24:44.440 --> 0:24:45.600 state of AI and music. 0:24:45.840 --> 0:24:47.920 I think I would say I'm quite both. One of 0:24:47.960 --> 0:24:49.600 the main things I'm doing right now is trying to 0:24:49.640 --> 0:24:52.600 figure out the ways to make this safe for humans, 0:24:52.880 --> 0:24:56.920 not even just existentially, but like socially and emotionally and civilizationally. 0:24:57.080 --> 0:24:58.760 But I also think that's totally possible. 0:24:59.080 --> 0:25:02.040