WEBVTT - The music label that changed its mind about AI

0:00:00.480 --> 0:00:03.720
<v Speaker 1>Already, and this is the Daily Art. This is the

0:00:03.800 --> 0:00:06.840
<v Speaker 1>Daily os. Oh, now it makes sense.

0:00:14.720 --> 0:00:17.400
<v Speaker 2>Good morning and welcome to the Daily Ours. It's Tuesday,

0:00:17.400 --> 0:00:19.760
<v Speaker 2>the fourth of November. I'm Emma Gillespie.

0:00:19.800 --> 0:00:20.759
<v Speaker 1>I'm Lucy Tassel.

0:00:21.120 --> 0:00:25.040
<v Speaker 2>One of the world's biggest music companies, Universal has announced

0:00:25.079 --> 0:00:28.400
<v Speaker 2>a major deal with an AI song generator, the first

0:00:28.440 --> 0:00:31.000
<v Speaker 2>of its kind partnership would have been a pretty big

0:00:31.000 --> 0:00:33.360
<v Speaker 2>global story on its own, but it's the fact that

0:00:33.479 --> 0:00:36.720
<v Speaker 2>Universal had been trying to sue the very platform it's

0:00:36.760 --> 0:00:41.240
<v Speaker 2>now working with that's added another layer of interest. Today

0:00:41.360 --> 0:00:43.680
<v Speaker 2>we are going to unpack how we got here and

0:00:43.720 --> 0:00:46.400
<v Speaker 2>what this deal means for the creative industry and the.

0:00:46.320 --> 0:00:52.680
<v Speaker 1>Future of music. Emma, to kick things off, I think

0:00:52.720 --> 0:00:54.600
<v Speaker 1>the best place to start is to explain who are

0:00:54.640 --> 0:00:57.160
<v Speaker 1>the key players in this story and how have they

0:00:57.240 --> 0:00:59.320
<v Speaker 1>interacted with each other up to this point. Yes.

0:00:59.400 --> 0:01:04.480
<v Speaker 2>So, first we've got Universal Music Group UMG, a massive,

0:01:04.680 --> 0:01:09.240
<v Speaker 2>massive company that owns labels including EMI, Capitol Records, def Jam,

0:01:09.400 --> 0:01:13.720
<v Speaker 2>Virgin Music. It calls itself the world's leading music company,

0:01:13.920 --> 0:01:18.760
<v Speaker 2>and for good reason. It represents artists including Taylor Swift,

0:01:18.840 --> 0:01:23.080
<v Speaker 2>Billie Eilish, Kendrick Lamar, Sabrina Carpenter, The Weekend, Dochy Drake,

0:01:23.200 --> 0:01:25.600
<v Speaker 2>Post Malone, you get the drift, I could go on,

0:01:26.080 --> 0:01:30.360
<v Speaker 2>then we've got Udio. This is an AI music generator.

0:01:30.400 --> 0:01:33.720
<v Speaker 2>So it launched in April last year with some backing

0:01:33.840 --> 0:01:37.160
<v Speaker 2>from a few quite big names in tech that includes

0:01:37.360 --> 0:01:42.800
<v Speaker 2>Will I Am the Musician and Instagram co founder Mike Krieger. Now,

0:01:43.000 --> 0:01:46.640
<v Speaker 2>how it works is users can type prompts into a

0:01:46.680 --> 0:01:51.200
<v Speaker 2>textbox on the platform and generate entirely new songs based

0:01:51.240 --> 0:01:54.640
<v Speaker 2>on those instructions. So you could, for example, type something

0:01:54.840 --> 0:01:58.600
<v Speaker 2>like upbeat song about podcasts and you'll get a song

0:01:58.880 --> 0:02:02.040
<v Speaker 2>pretty quickly or almost instantly. Or you can even get

0:02:02.080 --> 0:02:06.160
<v Speaker 2>more specific and request something that sounds like a particular

0:02:06.240 --> 0:02:10.280
<v Speaker 2>genre or style like, for example, upbeat pop song in

0:02:10.280 --> 0:02:13.359
<v Speaker 2>the style of Sabrina Carpenter about the Daily Os podcast.

0:02:13.680 --> 0:02:15.320
<v Speaker 2>And you can even go as far to kind of

0:02:15.360 --> 0:02:18.320
<v Speaker 2>describe the vibe that you want the song to kind

0:02:18.320 --> 0:02:20.760
<v Speaker 2>of have, how you want listeners to feel when they

0:02:20.800 --> 0:02:24.080
<v Speaker 2>hear it, and then the platform creates a track that

0:02:24.120 --> 0:02:25.000
<v Speaker 2>matches your description.

0:02:25.560 --> 0:02:29.639
<v Speaker 1>Okay, so this music platform, this AI music platform launchers

0:02:29.840 --> 0:02:32.400
<v Speaker 1>last year. You mentioned up at the top that at

0:02:32.440 --> 0:02:36.880
<v Speaker 1>some point, UMG decides to sue this platform, I assume

0:02:37.280 --> 0:02:40.760
<v Speaker 1>for the reasons that you've described. Basically it can mimic

0:02:40.880 --> 0:02:41.799
<v Speaker 1>its artists.

0:02:41.919 --> 0:02:42.800
<v Speaker 2>Yes, exactly.

0:02:42.880 --> 0:02:43.000
<v Speaker 1>So.

0:02:43.120 --> 0:02:47.120
<v Speaker 2>In June last year, the Recording Industry Association of America

0:02:47.240 --> 0:02:51.880
<v Speaker 2>announced Music's Big Three. So that's Universal, Sony and Warner

0:02:52.440 --> 0:02:55.639
<v Speaker 2>had teamed up to take action against Udio, that very

0:02:55.680 --> 0:02:59.880
<v Speaker 2>platform we just talked about, and another AI generator called Suno.

0:03:00.639 --> 0:03:05.239
<v Speaker 2>The music groups basically accused these platforms of copyright infringement

0:03:05.280 --> 0:03:09.000
<v Speaker 2>on what they called an almost unimaginable scale. The core

0:03:09.040 --> 0:03:12.799
<v Speaker 2>allegation was basically that these platforms were training their AI

0:03:12.880 --> 0:03:17.040
<v Speaker 2>models on millions of copyrighted songs without any permission or

0:03:17.120 --> 0:03:20.760
<v Speaker 2>licensing deals in place to do so. So the idea

0:03:20.840 --> 0:03:24.679
<v Speaker 2>being that these platforms are kind of profiting off other artists'

0:03:24.680 --> 0:03:26.799
<v Speaker 2>work without giving them their dues.

0:03:27.200 --> 0:03:29.560
<v Speaker 1>What kind of evidence did the Big Three present?

0:03:30.200 --> 0:03:35.520
<v Speaker 2>So Universal alleged that multiple AI generated songs on Udio

0:03:36.080 --> 0:03:41.640
<v Speaker 2>closely resembled their classics. It gave many examples, including a

0:03:41.720 --> 0:03:45.240
<v Speaker 2>song that had a very similar melody and chord progression

0:03:45.640 --> 0:03:49.320
<v Speaker 2>to the Temptations nineteen sixty four hit My Girl Now.

0:03:49.480 --> 0:03:52.400
<v Speaker 2>Udio generated a song that was very similar to My

0:03:52.520 --> 0:03:57.120
<v Speaker 2>Girl based on the prompt My Tempting nineteen sixty four girl,

0:03:57.360 --> 0:03:59.960
<v Speaker 2>Smokey sing Hitsville soul pop.

0:04:00.400 --> 0:04:00.920
<v Speaker 1>Interesting.

0:04:01.240 --> 0:04:05.480
<v Speaker 2>Yes, UMG pointed to other examples that sounded very similar

0:04:05.560 --> 0:04:09.400
<v Speaker 2>to classic hits like My Way as performed by Frank Sinatra,

0:04:09.880 --> 0:04:13.640
<v Speaker 2>Abbas Dancing Queen, and even Christmas classics like Rocking around

0:04:13.640 --> 0:04:17.200
<v Speaker 2>the Christmas Tree. So the labels argued that for Udio

0:04:17.279 --> 0:04:21.960
<v Speaker 2>specifically to create music that sounded this similar to existing songs,

0:04:22.400 --> 0:04:26.520
<v Speaker 2>it must have trained its model on copyrighted recordings. Then,

0:04:26.720 --> 0:04:30.080
<v Speaker 2>the record labels later amended the lawsuit to also include

0:04:30.120 --> 0:04:34.279
<v Speaker 2>allegations that Udio sourced music through the illegal scraping of

0:04:34.360 --> 0:04:37.520
<v Speaker 2>copyrighted recordings from YouTube as well.

0:04:37.920 --> 0:04:41.479
<v Speaker 1>Okay, right, So then how did Udio respond to these allegations.

0:04:41.880 --> 0:04:44.760
<v Speaker 2>It argued that it hadn't broken any laws and that

0:04:44.839 --> 0:04:47.880
<v Speaker 2>it uses quote state of the art filters to ensure

0:04:47.880 --> 0:04:52.320
<v Speaker 2>it does not reproduce copyrighted works or artists' voices. The

0:04:52.320 --> 0:04:57.520
<v Speaker 2>company claims its music generator was trained using publicly available music. Okay,

0:04:57.560 --> 0:05:01.760
<v Speaker 2>and this has started a really interesting conversation about fair use.

0:05:02.000 --> 0:05:05.039
<v Speaker 2>So it said its actions constituted fair use under US

0:05:05.160 --> 0:05:09.640
<v Speaker 2>copyright law. Which is a legal concept that basically allows

0:05:09.800 --> 0:05:13.960
<v Speaker 2>a limited use of some copyrighted material without permission in

0:05:14.080 --> 0:05:18.720
<v Speaker 2>certain circumstances. It's quite vague, but an example of fair

0:05:18.839 --> 0:05:21.320
<v Speaker 2>use might be if we hear on the podcast we're

0:05:21.320 --> 0:05:24.760
<v Speaker 2>talking about a story relating to a popular song, and

0:05:24.800 --> 0:05:27.080
<v Speaker 2>we played, you know, five seconds of that song to

0:05:27.120 --> 0:05:31.080
<v Speaker 2>set the scene, that would be considered fair use. Udio said, quote,

0:05:31.120 --> 0:05:33.960
<v Speaker 2>we stand behind our technology and believe that generative AI

0:05:34.120 --> 0:05:36.479
<v Speaker 2>will become a mainstay of modern society.

0:05:36.960 --> 0:05:39.760
<v Speaker 1>That's interesting that they said publicly available. They didn't say

0:05:39.920 --> 0:05:44.400
<v Speaker 1>royalty free, they didn't say like copyright free. They just

0:05:44.440 --> 0:05:47.000
<v Speaker 1>said publicly available, which I guess you could say. YouTube

0:05:47.040 --> 0:05:49.200
<v Speaker 1>is a platform that doesn't require you to sign in,

0:05:49.320 --> 0:05:51.960
<v Speaker 1>so you could kind of argue this is publicly available.

0:05:52.080 --> 0:05:55.280
<v Speaker 2>Yeah, a very interesting kind of rabbit hole of what

0:05:55.320 --> 0:05:56.120
<v Speaker 2>that really means.

0:05:56.640 --> 0:05:58.640
<v Speaker 1>So then how do we get from there this kind

0:05:58.680 --> 0:06:02.280
<v Speaker 1>of legal standoff to a business partnership.

0:06:02.720 --> 0:06:05.480
<v Speaker 2>We now know that the two companies had actually been

0:06:05.480 --> 0:06:08.280
<v Speaker 2>in talks behind the scenes for several months while these

0:06:08.320 --> 0:06:11.560
<v Speaker 2>ongoing legal proceedings were in the background, and it all

0:06:11.600 --> 0:06:14.520
<v Speaker 2>came to a head last week when Udio and Universal

0:06:14.600 --> 0:06:19.520
<v Speaker 2>confirmed they had reached a compensatory legal settlement, but it

0:06:19.560 --> 0:06:22.400
<v Speaker 2>didn't stop there, because the companies announced that they're now

0:06:22.520 --> 0:06:26.479
<v Speaker 2>partnering to build an entirely new AI music platform together,

0:06:26.880 --> 0:06:29.000
<v Speaker 2>which is set to launch in twenty twenty six.

0:06:29.360 --> 0:06:30.960
<v Speaker 1>What do we know about this platform?

0:06:31.120 --> 0:06:34.760
<v Speaker 2>According to a joint statement from both companies, it's going

0:06:34.800 --> 0:06:38.360
<v Speaker 2>to be powered by Generative AI, trained on authorized and

0:06:38.640 --> 0:06:43.600
<v Speaker 2>licensed music, and the company said a new subscription service

0:06:43.680 --> 0:06:47.360
<v Speaker 2>will be involved, so users will pay subscribe to kind

0:06:47.360 --> 0:06:51.000
<v Speaker 2>of use the platform that is based off accessing this

0:06:51.240 --> 0:06:55.160
<v Speaker 2>licensed music, which the statement said will quote transform the

0:06:55.279 --> 0:06:59.240
<v Speaker 2>user engagement experience, creating a licensed and protected environment to

0:06:59.320 --> 0:07:03.440
<v Speaker 2>customize and share music responsibly on the Udio platform. A

0:07:03.480 --> 0:07:07.599
<v Speaker 2>crucial difference from the old Udio is that this platform

0:07:07.640 --> 0:07:11.120
<v Speaker 2>will only work with music from artists who have opted

0:07:11.320 --> 0:07:14.320
<v Speaker 2>into it, who have given their permission explicitly.

0:07:14.560 --> 0:07:17.040
<v Speaker 1>Okay, yeah, so that's the big difference. It's not taking

0:07:17.160 --> 0:07:21.600
<v Speaker 1>things from YouTube. It's saying, I, an artist, choose to

0:07:21.680 --> 0:07:24.840
<v Speaker 1>allow this AI platform to use my songs to come

0:07:24.920 --> 0:07:25.840
<v Speaker 1>up with new material.

0:07:26.080 --> 0:07:29.640
<v Speaker 2>Yes, And I mean that's what Universal says, That's what

0:07:29.840 --> 0:07:33.640
<v Speaker 2>Udio says. They're framing it as this kind of further

0:07:33.800 --> 0:07:38.440
<v Speaker 2>revenue opportunity for UMG artists and songwriters. But in terms

0:07:38.480 --> 0:07:41.480
<v Speaker 2>of what it actually looks like, there's a lot that

0:07:41.560 --> 0:07:45.320
<v Speaker 2>we don't know. It could mean, and it likely will

0:07:45.320 --> 0:07:49.000
<v Speaker 2>mean that udio users can generate AI songs using the

0:07:49.120 --> 0:07:53.200
<v Speaker 2>actual voices of well known artists. The CEO of UDO,

0:07:53.240 --> 0:07:56.200
<v Speaker 2>a guy called Andrew Sanchez. He told Rolling Stone, quote,

0:07:56.320 --> 0:07:59.080
<v Speaker 2>we want to build a community of super fans around creation.

0:07:59.560 --> 0:08:01.840
<v Speaker 2>You'll be able to consume and interact with your favorite

0:08:01.840 --> 0:08:04.960
<v Speaker 2>songs and artists in the same place, adding I might

0:08:05.000 --> 0:08:06.480
<v Speaker 2>want to listen to songs that are made by my

0:08:06.520 --> 0:08:09.160
<v Speaker 2>favorite band, and then maybe I want to remix one.

0:08:09.520 --> 0:08:13.000
<v Speaker 2>He called it a massive expansion, a paradigm that doesn't

0:08:13.080 --> 0:08:13.920
<v Speaker 2>exist right now.

0:08:14.480 --> 0:08:18.680
<v Speaker 1>So these artists, these songwriters, they're opting in to this platform,

0:08:18.760 --> 0:08:20.320
<v Speaker 1>are they going to get paid for that use?

0:08:20.680 --> 0:08:23.320
<v Speaker 2>Well, that is what Universal says. It says there will

0:08:23.320 --> 0:08:27.240
<v Speaker 2>be revenue opportunities for artists that opt in. It looks

0:08:27.280 --> 0:08:30.440
<v Speaker 2>like they will be compensated in two ways, So when

0:08:30.480 --> 0:08:33.720
<v Speaker 2>their music is used to train the AI model, and

0:08:33.840 --> 0:08:37.280
<v Speaker 2>also when their music is used in the platform's outputs.

0:08:37.440 --> 0:08:40.280
<v Speaker 2>So if a song uses the voice of an artist

0:08:40.360 --> 0:08:43.559
<v Speaker 2>or sounds like an artist because it's trained on their songs,

0:08:43.800 --> 0:08:46.360
<v Speaker 2>they're kind of product of that will include some kind

0:08:46.360 --> 0:08:49.160
<v Speaker 2>of compensation for artists. We don't know a great deal

0:08:49.200 --> 0:08:52.640
<v Speaker 2>about how much money will be in it for the artists,

0:08:52.920 --> 0:08:56.080
<v Speaker 2>but we also know that there is this bigger compensation

0:08:56.200 --> 0:08:59.599
<v Speaker 2>deal that has been reached between UMG and Udio. The

0:08:59.640 --> 0:09:03.600
<v Speaker 2>exact figure of what Udio has agreed to pay Universal, though,

0:09:03.840 --> 0:09:04.959
<v Speaker 2>has not been made public.

0:09:05.440 --> 0:09:08.080
<v Speaker 1>All of these changes, though, are still in the future.

0:09:08.160 --> 0:09:10.920
<v Speaker 1>I assume what happens to Udio in interim.

0:09:11.400 --> 0:09:15.240
<v Speaker 2>In the meantime, the platform has already been updated since

0:09:15.280 --> 0:09:18.559
<v Speaker 2>this deal was announced so that users can still create

0:09:18.640 --> 0:09:23.160
<v Speaker 2>AI generated songs and stream within Udio, but they can

0:09:23.200 --> 0:09:26.960
<v Speaker 2>no longer download tracks or share them on other platforms

0:09:27.000 --> 0:09:30.320
<v Speaker 2>like Spotify or Apple Music or YouTube, which was an

0:09:30.360 --> 0:09:32.680
<v Speaker 2>early challenge of AI soong generation that you know, people

0:09:32.720 --> 0:09:36.000
<v Speaker 2>were making music that sounded a lot like the artists

0:09:36.040 --> 0:09:38.680
<v Speaker 2>we know and listen to on a regular basis, sharing

0:09:38.679 --> 0:09:41.800
<v Speaker 2>them elsewhere, but them being you know, not the real deal,

0:09:41.960 --> 0:09:45.280
<v Speaker 2>and they're being lost revenue for those artists. But as

0:09:45.280 --> 0:09:47.840
<v Speaker 2>I mentioned, as soon as this deal was announced last week,

0:09:48.040 --> 0:09:51.920
<v Speaker 2>Udio immediately disabled all downloads from its platform.

0:09:52.000 --> 0:09:55.280
<v Speaker 1>Speaking of those artists. What have we heard from singers

0:09:55.280 --> 0:09:56.280
<v Speaker 1>and songwriters so far?

0:09:56.760 --> 0:10:00.000
<v Speaker 2>Yeah, the response within the music industry has been pretty mixed.

0:10:00.720 --> 0:10:03.560
<v Speaker 2>Some see it as a positive step forward, others have

0:10:03.640 --> 0:10:09.360
<v Speaker 2>raised serious concerns. The Music Artists Coalition, which represents recording artists,

0:10:09.440 --> 0:10:13.480
<v Speaker 2>released a statement raising questions around the deal, and they

0:10:13.520 --> 0:10:17.040
<v Speaker 2>are asking for more transparency around some key issues, you know,

0:10:17.120 --> 0:10:20.440
<v Speaker 2>things like how much settlement money was paid, where will

0:10:20.440 --> 0:10:23.800
<v Speaker 2>this money go, how will revenue be split, and how

0:10:23.840 --> 0:10:27.320
<v Speaker 2>will artists work actually be used in practice if they

0:10:27.360 --> 0:10:30.840
<v Speaker 2>give consent to Udio. Then we've got us based not

0:10:30.920 --> 0:10:34.199
<v Speaker 2>for profit, the Copyright Alliance. It represents over two million

0:10:34.320 --> 0:10:37.480
<v Speaker 2>artists and creatives and has actually welcomed the deal. It

0:10:37.559 --> 0:10:40.760
<v Speaker 2>said this will create a new, responsibly trained, generative AI

0:10:40.840 --> 0:10:41.559
<v Speaker 2>music service.

0:10:41.920 --> 0:10:42.280
<v Speaker 1>Quote.

0:10:42.320 --> 0:10:44.360
<v Speaker 2>This news isn't just a win for UMG and for

0:10:44.400 --> 0:10:47.120
<v Speaker 2>its artists and songwriters, it's a win for everyone in

0:10:47.120 --> 0:10:50.840
<v Speaker 2>the copyright community. It reaffirms that generative AI products should

0:10:50.840 --> 0:10:54.920
<v Speaker 2>and can be crafted responsibly and with respect for copyright owners.

0:10:55.000 --> 0:10:56.720
<v Speaker 2>So that was from the Copyright Alliance.

0:10:57.280 --> 0:11:02.560
<v Speaker 1>Now I am assuming this has resolved and Udio's legal battle,

0:11:02.559 --> 0:11:04.360
<v Speaker 1>But what about the other two in the big three

0:11:04.400 --> 0:11:05.400
<v Speaker 1>that you mentioned earlier.

0:11:05.600 --> 0:11:08.800
<v Speaker 2>Yes, Sony and Warner are still actively pursuing this lawsuit

0:11:08.920 --> 0:11:11.640
<v Speaker 2>with Udio, so the platform is still being sued by

0:11:11.640 --> 0:11:15.280
<v Speaker 2>two of the three big players. But the UMG deal

0:11:15.520 --> 0:11:19.000
<v Speaker 2>really could mark a turning point for AI and music.

0:11:19.040 --> 0:11:22.440
<v Speaker 2>You know, we could see potentially Sony and Warner follow

0:11:22.520 --> 0:11:26.320
<v Speaker 2>suit and strike similar licensing deals. But there are some

0:11:26.440 --> 0:11:29.800
<v Speaker 2>big questions that remain, you know, like the question around

0:11:29.800 --> 0:11:33.320
<v Speaker 2>what is fair use. Can AI companies legally train their

0:11:33.360 --> 0:11:37.720
<v Speaker 2>models on copyrighted music without authorization? We have no answer

0:11:37.720 --> 0:11:41.360
<v Speaker 2>to that. If the Sony and Warner lawsuits proceed, it

0:11:41.440 --> 0:11:43.440
<v Speaker 2>will be you know, really interesting to see what a

0:11:43.520 --> 0:11:46.760
<v Speaker 2>judge thinks about that. And you know, how the UMG

0:11:46.840 --> 0:11:49.280
<v Speaker 2>approved Udio platform of the future is going to work

0:11:49.280 --> 0:11:53.240
<v Speaker 2>in practice for artists and users is a massive question mark.

0:11:53.440 --> 0:11:55.480
<v Speaker 2>Whatever happens next, though, Lucy, it's clear this is a

0:11:55.520 --> 0:11:58.520
<v Speaker 2>deal that's going to shape how the music industry approaches

0:11:58.600 --> 0:12:01.560
<v Speaker 2>AI going forward, and we'll keep a close eye on it.

0:12:01.840 --> 0:12:03.719
<v Speaker 1>Thanks so much for explaining that to us, and what

0:12:03.880 --> 0:12:06.560
<v Speaker 1>Thanks Lucy, and thank you for joining us today on

0:12:06.679 --> 0:12:09.599
<v Speaker 1>the Daily Odds. We'll be back this evening with the headlines.

0:12:09.720 --> 0:12:16.040
<v Speaker 1>Until then, have a great day. My name is Lily

0:12:16.040 --> 0:12:19.439
<v Speaker 1>Maddon and I'm a proud Arunda Bungelung Kalkudin woman from

0:12:19.480 --> 0:12:23.840
<v Speaker 1>Gadigol Country. The Daily oz acknowledges that this podcast is

0:12:23.880 --> 0:12:26.360
<v Speaker 1>recorded on the lands of the Gadighl people and pays

0:12:26.400 --> 0:12:29.600
<v Speaker 1>respect to all Aboriginal and Torres Strait island and nations.

0:12:29.920 --> 0:12:32.840
<v Speaker 1>We pay our respects to the first peoples of these countries,

0:12:32.960 --> 0:12:34.120
<v Speaker 1>both past and present.