- il y a 1 jour
Tuning Into the Future: How is AI Reshaping the Music Industry
Catégorie
🤖
TechnologieTranscription
00:00Bonjour à tous, merci beaucoup d'être ici avec nous aujourd'hui.
00:05Je suis très heureux de parler de l'AI et de la musique aujourd'hui avec vous.
00:11Avec me sur le stage nous avons Anthony Browns,
00:14votre Composer et Professeur de Composition et Theory à Rice University,
00:19Nathalie Birochot, votre CEO de Arcam Amplify,
00:22et Manuel Moussalam, votre Director de Research à Deezer.
00:27Nous avons beaucoup de topics à parler aujourd'hui.
00:31Toutes les industries de créativité sont déjà impactées par l'AI aujourd'hui.
00:37Je vais commencer par vous, Manuel.
00:40Comment est-ce que l'on utilise l'AI aujourd'hui ?
00:44Comment est-ce que l'on utilise l'AI aujourd'hui ?
00:47On l'aise, nous avons utilisé l'AI et l'enseignement de l'enseignement de l'enseignement de l'enseignement.
00:51Je travaille dans le département de l'enseignement de l'enseignement,
00:55donc nous faisons des recherches sur les networks,
00:57sur ces types de technologies pour un temps.
00:59Nous essayons de l'utiliser pour l'enseignement de l'enseignement de l'enseignement.
01:03Mais l'avenir, le changement de l'enseignement a été l'utiliser des technologies de l'enseignement,
01:07comme les modèles de l'enseignement de l'enseignement de l'enseignement,
01:18électroniquement et délivés à notre plateforme.
01:20Donc, nous voyons une massive increase
01:22dans l'enseignement de l'enseignement de l'enseignement de l'enseignement
01:25et c'est la main change que nous sommes en train de l'enseignement.
01:30Et pour vous, des questions, Nathalie ?
01:32Merci.
01:33Et hello, everyone.
01:34So happy to be there this afternoon.
01:37Just to clarify,
01:38just at the beginning,
01:39What is YoCam Amplify ?
01:40We are a tech company
01:42and we provide technology and intelligence
01:45to companies in the music industry
01:48like Deezer and other music platforms
01:50of labels, distributors and DSPs.
01:53And our mission is to enrich and protect
01:55the human creation for the future.
01:58So, what does it mean ?
01:59We empower creators with technology
02:01and we secure the industry to deliver
02:03a high-end sound experience for the future.
02:08That being said, maybe you know IRCAM.
02:11So, we are a spin-off of IRCAM.
02:13IRCAM is a famous lab in France
02:16that's existing for 50 years.
02:19And so, having said that,
02:21to answer the question,
02:25today we develop an AI music detector.
02:28So, we have a tool to detect
02:29if a track, if a music has been made with AI.
02:33And what we can say,
02:35as Deezer said, as Manuel said,
02:38it's that there has been a huge invasion
02:45of the music distribution platforms
02:47by millions and millions of AI produced tracks.
02:51So, it is important to detect them,
02:53to be more intelligent collectively
02:56and to take the right decision for the future.
02:59And just one more precision,
03:02and after that I let Dominic to Anthony.
03:06The other problem is that
03:08the streaming of these songs
03:10is manipulated, you know,
03:13to create false stream in a fraudulent way.
03:17So, this is another problem.
03:19And we think that such neutral technology
03:23as IRCAM can help the whole ecosystem
03:26to bring value.
03:28And for you, Anthony,
03:30how is AI transforming the composer's creative process?
03:34Do you see it like an opportunity or a threat?
03:39So, it's exciting to be at Vivitech
03:41among so much amazing technology.
03:43And I've been thinking to myself,
03:44we have Silicon Valley,
03:45but you have Silicon Continent.
03:47It's incredible.
03:48And it's an honor to be here,
03:50in a way, representing the arts in this milieu.
03:53What I would say is that,
03:55right now, AI is extremely good
03:57at what I would call disposable art.
03:59It's almost like single-use plastic.
04:01And I want to make a plea
04:04that the technology of the future
04:06incorporate artists in its development.
04:09I think creativity is an inherently,
04:11massively interdisciplinary challenge.
04:14And anything having to do with the arts
04:16should involve artists.
04:18So, I see incredible opportunity,
04:19but in what Natalie is representing.
04:23And one of my heroes is Billy Kluber
04:25from Bell Labs in the 1960s
04:27who worked with Robert Rauschenberg
04:29and John Cage and Merce Kenningham
04:31and Lucinda Childs,
04:32and created this interdisciplinary model,
04:34which I think needs to be revived
04:36for the age of AI.
04:38At this year, how many, I don't know,
04:41of the percentage of music generated by AI at this year?
04:45Okay, so first,
04:47to be really clear on what we detect,
04:50we only detect stuff that has been
04:52100% generated by Suno and Udio
04:55and this kind of model.
04:57No further manipulation,
04:59no creative process whatsoever.
05:00It's really like the raw output of this kind
05:02of generative models.
05:04So, what we deem as really low quality value
05:08for our customers.
05:10If we stick to that,
05:11we already get to a pretty impressive number.
05:15So, we identify more than 20,000 tracks every day
05:19being delivered to a platform,
05:21which amounts to a little more than 18%
05:24of everything that is delivered on the platform.
05:27So, that's 18% content that down the line
05:31gets almost zero stream.
05:33So, no one is actively listening to this kind of content,
05:36apart from some fraudulent activity, as you mentioned.
05:39Indeed, it's one of the things we also observe.
05:43So, this machine generated content is mostly machine consumed,
05:48also, which is not so interesting for us, really.
05:52And it's a bit...
05:53The risk is dilution of revenues for artists,
05:57and this is really what we want to prevent.
05:58And also, from a very practical technical standpoint,
06:02because I'm a technical guy here,
06:05it's a mess to store all this file to...
06:08I mean, we have to process everything.
06:10It's a lot of computation.
06:11It's a lot of handling the waste,
06:13the disposable plastic, as you mentioned.
06:15It's a nice image.
06:17I like it.
06:18It's a lot of work for us.
06:21You want to...
06:22Yeah, I think one of the issues, too,
06:24is it changes the risk model for artists.
06:27I mean, creativity, as everyone in these industries know,
06:30is inherently risky.
06:32And you think of, let's say, Picasso in the early 20th century,
06:36he's in an artist garret,
06:37and maybe there's 40 other people also in garrets.
06:40And they're all taking a risk,
06:42and Picasso has to worry about the careers of the 39 other artists.
06:47Once there's a million, or 10 million, or 20 million,
06:51and at the push of a button, you're generating content,
06:54and now Picasso is worried about all of that competition
06:56and trying to stand out, the risk model changes.
06:59And then the question, how much can he afford to pursue these interesting
07:05and unusual and edge-case kinds of ideas,
07:08when the chance of being found get more and more difficult?
07:14That's another issue that what Manuel is talking about brings up.
07:18And what difference do you see between human creativity and, let's say, algorithmic creativity?
07:27Maybe Natalie?
07:30It's like an easy one first, because at IRCAM, and we are a spin-off of IRCAM,
07:35and at IRCAM there are scientists from CNRS,
07:37Urban University for 50 years,
07:39and they are working on the link between technology and algorithms and human creativity.
07:46This is like the core of the research.
07:49So we are working on the link between the human creation and the use of the technology.
07:55And we talk about artificial intelligence, but because of Gen AI, the stakes are really high today,
08:02because of volume and speed.
08:04But artificial intelligence has been existing in IRCAM for like 10 years or 16 years.
08:09And what we are working on today with that AI detector and other works and other projects we are working
08:18on,
08:18is to give some very concrete tools to help the creative music industry to build new models
08:27and to have the tools and to have the data and the intelligence to build the right models
08:33for the right monetization, for the human creation, and to tackle the different challenges,
08:40and to be able to have the knowledge, you know?
08:44And I heard when we were just right there a few minutes ago,
08:49to know the future you need to know the truth.
08:53It's a little bit like this. You need the details, you need the intelligence.
08:56So that's why we think that neutral tools can help the different players in the value chain,
09:03so it just returns the DSPs, the rights management companies, to have the data and the metadata
09:09to build the right models. And that's very important.
09:15If I can add maybe on your question, I just wanted to say everything is human generated.
09:20I mean, engineers are human, trust me.
09:24So the people who build these machines are all humans.
09:27And most importantly, these generative models, they are trained on human material.
09:31That's why they create music that sounds like human music.
09:36It's because all this human work is being leveraged.
09:39So it's more a matter of who gets to extract the value and exploit it,
09:44rather than a human versus machine in this spirit.
09:49I'd say there are two very important differences between human and AI creativity.
09:54Since the earliest days of artificial intelligence, AI creativity has been based on randomness.
10:01And human creativity is not based on randomness.
10:05The great neuroscientist Anna Abraham has this interesting example of the basketball player,
10:10Kobe Bryant, who took ballet classes and incorporated lessons that he learned in ballet for moves
10:17that surprised his opponents in the middle of a basketball game.
10:20That's not being randomly creative.
10:22That's being multimodal and transdisciplinary and crossing domains and mixing things together.
10:28Human brains do that fantastically.
10:30AI doesn't see creativity through that lens.
10:34The other interesting paradox is that AI is trying to be creative by aiming for mainstream solutions.
10:41It's largely trained to aim for consensus because more than anything it wants to be right.
10:48It wants to be successful.
10:49And humans, of course, have this amazing ability to go to the outlier and the edge cases and the unusual.
10:55And they do something I call amplifying the anomaly, which is like world building.
11:01Out of their most original ideas, they manage to develop and make the most of them.
11:07AI, as currently designed, cannot see creativity that way.
11:12So those are two very important differences.
11:15Yeah. Creativity is all about mistakes and errors in the iterative process.
11:21So that's so true. I really agree with that.
11:25And yeah, no, it's okay.
11:27What are the main challenges, you know, regarding AI for players, maybe like you, the platform for you and then
11:36for, of course, for the artists and maybe for the labels?
11:41Well, so I'm a researcher. I lead a team of researchers.
11:45So the main challenges for us are work opportunities.
11:48So I can list a lot.
11:50But basically, we haven't solved AI detection.
11:54That's how we see the problem.
11:57We've published a paper recently, and we're going to publish another one in an upcoming conference, where we tried to
12:04beat our own detector.
12:05And we found it was excessively easy to beat our own detector, which is not great news for us.
12:13So basically, for me, the challenges are almost the same as they were six months ago.
12:19We haven't really found a way to detect all the AI-generated content that is out there.
12:26We detect some amount of it. It's already huge.
12:30But we still need to work. We haven't finished our work.
12:35Okay, for you, Natalie.
12:39First, just let's say that AI is, I don't see AI as a challenge, because it's the object of the
12:48research and of the thing we're developing and the basis of our product.
12:52So it's more like, okay, this is a material, a new opportunity to create, a new opportunity to imagine, to
13:03push the boundaries of creation, of detection, of protection of different catalogs,
13:09a new way of improving the productivity, etc., etc., you know that.
13:18Technically, this is a cat-and-mouse play, you know.
13:23When we talk about AI detection, of course, we run after something, we run after the content that are created
13:33by the solutions.
13:34And to be always up-to-date, it's important to have the connection with all the ecosystem and the feedbacks
13:43from the market,
13:44because the challenges are not the same. It's not the same point of view when you're a streaming platform or
13:51a rights management society or an artist or an aggregator.
13:57It's not the same challenges.
13:59And the main challenge, according to me, is to find a way to put solutions in the market that could
14:08be helpful and very concrete as a first step, because the longest road.
14:13And as a first step, just to help, to be the first prick to help.
14:19And on the technical aspect, what we can see when we are talking notably with the rights management company is
14:28that they need more information.
14:31So, they need the information to be reliable and to be deeper.
14:36So, like, is it AI per stem?
14:42So, the violin is the violin AI, is the voice AI, etc., etc., etc.
14:50This part is AI and this part is AI.
14:53The confidence score, is it reliable, is it more and more and more reliable and more and more strong?
15:00The false positive rate, so we work on it a lot to decrease the rate of false positive.
15:09It's like, of course, when you've got tens and tens of millions of strikes to process, you need to be
15:16really smart on that dimension.
15:19And so, the stakes are high, the challenges are great, big, but, yeah, that's it.
15:29Thank you.
15:30And I think for artists, there are two challenges I would highlight.
15:34I read recently that Cezanne took a hundred sessions to paint a still life and a hundred and fifty sessions
15:42to paint a portrait.
15:44And so, you ask yourself, well, why is he spending so much time doing that?
15:48And because he's creating art with shelf life.
15:51He's creating art that a hundred years later, people still want to look at.
15:56And I worry about a world in which, by entering a few text prompts, AI generates your work for you.
16:05The idea of spending four to six months laboring over something and making all the thousands of choices it takes
16:11to make something creative will seem like a massive waste of time.
16:14But it will also marginalize voices humanity has always enjoyed hearing from.
16:20And, in fact, we treasure and we pass down to the future.
16:23So, that's one risk.
16:25And the second is, we know perfectly well from history that there are works of art that enter the cultural
16:31bloodstream but aren't accepted right away.
16:34But then, maybe a generation later, they turn out to be the best of what was made.
16:38And, of course, reinforcement learning right now is used a lot to curate AI products, but it's based on, oh,
16:45do I like it right now?
16:46No one brings it back a year later and says, do you still like it?
16:51Is it still worth listening to or looking at it?
16:53We don't yet have a formula for dealing with the longer trajectory of the acceptance of art.
16:59And, without that, we risk creating a whole new type of culture where it's all about, well, it feels good
17:06today and I'll just make something else tomorrow.
17:09And we won't have the kind of records and testaments and, you know, deep thought that we have traditionally cherished
17:17as a society.
17:18Well, I agree.
17:19I think it's, I've been doing a lot of much smaller intervention lately on AI.
17:25And the question everyone asks is whether people are going to switch from a model where you listen to work
17:31of art and you build a relationship with them and it stays in your life.
17:36And, yeah, it has a meaning and it stays in the long term to a model where you just create
17:43disposable music that you consume because it's cheap and it's there and it's free.
17:49I personally don't think we're going to switch to that model.
17:52And, frankly, if you look at, if we look at numbers at Deezer, we have millions of customers and I
17:59don't see any trend of people starting listening to this content massively.
18:03I don't think it's, there is a niche emerging in this, but maybe I'm mistaken.
18:08But that's, yeah, that's the main question I think that everyone asks and I think it's too soon to answer.
18:15I just, personally, I don't want it to be true, so I certainly hope it's not going to go that
18:23way.
18:23And, Anthony, do you think AI would democratize the musical creation?
18:29Or, in the contrary, it will increase, you know, the inequalities between the artists and the labels?
18:36I mean, again, there's opportunity and there's challenges.
18:39I mean, it takes a nuclear power plant to run some of the servers that are in charge of ChatGPT
18:48and the large language models.
18:50And, if you're an individual artist, to stand up against the kind of a monolith that it takes to manufacture
18:57and control the data that's producing this machinery, it's very daunting.
19:05And, yes, there's an incredible opportunity for people to have entry level into creativity.
19:10But we also know that there are these subscription models in a lot of these services.
19:15And the quality of the data is a lot determines the quality of the output.
19:20And if you have your lower tier and then your gold tier and your platinum tier and the platinum tier
19:25is only available to people with means, then it won't democratize it at all.
19:29So, I think the challenge here at VivaTech is to make this as open access as possible, to find ways
19:37sustainably, economically, to truly offer opportunity to everyone.
19:42And I think that that would be a very noble enterprise.
19:46So, I want to talk about artists' rights.
19:49So, what measure could be put in place to protect artists' rights with all this generative content?
19:58Well, we don't have anyone from a generative AI company on stage with us.
20:03But maybe first license the content before you train your system.
20:07That would be a first step.
20:08I don't want to...
20:09I mean, that's a big next step, obviously.
20:13Right now, artists are not being paid for their work to be used, again, commercially by these companies.
20:21And I think that's a problem.
20:24There are a lot of considerations on the legal aspect.
20:27Whether they have the right to do it or not is still a matter of legal questions.
20:34But I think, yeah, I mean, again, I think it's not about machines.
20:39Computers are impressive and it takes a lot of energy to do that.
20:43But all they do is computations.
20:45And it's all just computation.
20:47And it's not thinking.
20:49It's not whatever us humans are doing.
20:52It's just a few computations on machines.
20:56Everything we see, even when it's generated by Suno, is still, somehow, the byproduct of human creation.
21:02And this has to be recognized and we have to acknowledge that even if it comes out of a LLM
21:08or a regenerative model, there's still a lot of human work behind it.
21:12And that work needs to be recognized.
21:14And I think the most we can do for artists is to emphasize that fact and say that it's not
21:20just a machine product.
21:21It comes from their work.
21:22And if it comes from their work, they should be compensated.
21:27What I can see, and if I can say something about artists, but artists are not our direct clients, but
21:34there is a huge aspect of education in the market.
21:39And we can observe that the level of maturity is really different from someone to another one in the artist
21:50area and in the music industry ecosystem.
21:54And that's difficult because it's going very, very, very fast.
21:58And so it's important to make the education, to have this type of conference, to have the data, to have
22:09the knowledge to explain what is happening.
22:13And what is doing by DISA with the communication and all the transparency about the data.
22:19Yes, it's so important because it's huge, it's now, it's fast.
22:23And the industry and the models were not prepared to that.
22:29They are based on models that have been defined after the MP3 crisis in 1999.
22:37Well, it's now.
22:39And according to me, it's like a little vertiginous to think about that.
22:46And I don't know what you think, Anthony, but alongside the community of artists, this is like, there is a
22:53big difference in the maturity concerning artificial intelligence.
23:00Yeah, and I would say, I would love it to be an industry best practice always to have artists partnered
23:07with the tech.
23:09And that the more you have a true alliance, the healthier the outcomes are going to be for everybody involved.
23:16And in a way, I think the products will be more interesting.
23:19It will benefit artists rather than compete with them.
23:22And I think ultimately it will have more staying power and lead to really, truly wonderful innovations that people will
23:29appreciate.
23:29And I think that it's a great challenge to artists right now to bring out the best of their inventiveness
23:38and to surpass the machinery in terms of the excitement that what we make generates.
23:45And I can imagine a future, right now you go see a, you know, hear a Beethoven symphony or a
23:51Tchaikovsky symphony.
23:51In the future, it may be you go hear the work of a composer and an engineer, and they're credited
23:56equally as co-creators of the work.
24:00That would be a fantastic thing.
24:01But it depends on really a long-term vision and a best practice of always involving artists.
24:08Yeah, Manuel, how do you see the collaboration in the future between musicians, researchers, and engineers?
24:16Well, I've been also trained at IRCAM during my studies.
24:20It's an amazing place where engineers and artists collaborate for more than half a century.
24:26So this is exactly the kind of initiatives.
24:29And I agree completely with Anthony also.
24:31It's everything that brings people together and brings creativity in the process.
24:37And I believe that machines are fantastic musical instruments.
24:40And I really think that.
24:42And when they are used as musical instruments by engineers and musicians, we're going to see a lot of very
24:49interesting musical pieces.
24:51It's just not, it's not cheap.
24:55I mean, you know, it's not that cheap.
24:56I mean, if you just create hundreds of thousands musical tracks with your computer in a few seconds, it's not
25:03going to be interesting.
25:04It's never going to interest anyone.
25:06And it's not what making music is about.
25:09So I truly believe this technology can be used to create very interesting stuff by people who take time, take
25:18time to see what they can do with it, try stuff, fail.
25:24Failure is extremely important.
25:26And then iterates.
25:28But please, just a few songs, not the tens of thousands that we get on a daily basis.
25:33We don't want that.
25:34No one wants that.
25:36Nathalie, if you want to hear something.
25:38Yeah, that's a funny one.
25:40Yeah, this is like philosophy.
25:43What is art?
25:44What is creation?
25:47But we are in an era where technology will be more and more powerful and the models and the interaction
25:58and the interface between human and machine is in an interesting period.
26:08And as I said, yes, in IRCAM, this is the beginning of IRCAM and Pierre Boulez put in the same
26:13place artists, scientists and engineers 50 years ago to say,
26:18well, we are in 77 and there is something new on earth, which is called computers.
26:28So at the real beginning of computing and algorithm, he thinks, and it was a genius, that the creation of
26:38music will be transformed by machines, by computers.
26:43And we are now 50 years later, and it's another cycle.
26:48It's another chapter of that adventure.
26:51And on the side of IRCAM Amplified, so the commercial arm of IRCAM, we wish this to be on the
27:01side of the different players in the value chain of music industry and on the side of artists to help
27:09to define the new rules.
27:11And we don't, we won't define the rules, but we on the side of the different players to help to
27:19give the tools and the intelligence to define the rules and to build, I hope, a world where the human
27:26creation will be well valued, recognized,
27:29and, yeah, to assure that there is a fair remuneration and, at the end, a balance which will be fair
27:43between the poor content and a work, an artistic work.
27:51And that's so important for all of us.
27:55Yeah, I love what Manuel and Natalie are saying.
27:58And there was an article in the New York Times earlier this week about a startup that's trying to take
28:05away everyone's job.
28:06And I think there's sort of a fork in the road.
28:09I imagine so much of the AI here goes to a hospital and says, what do you need help with?
28:16Goes to different industries and say, how can what we're doing support you?
28:22Do the same with the arts.
28:24You know, a partnership between the artists and the technology is going to be amazing.
28:28And the stuff that Natalie is involved in is incredible.
28:31And I was a fellow at IRCOM in 1984, so I, again, speak to how far-seeing it was and
28:38what it was doing.
28:39But if the goal is to take away everyone's job, including that of artists, I think humanity is going to
28:45suffer.
28:47So, Manuel, are you more optimistic for the future, for the industry, with AI?
28:56On some aspects, I'm always optimistic.
28:58So, yeah, I think we have, I believe in science.
29:02I think we have a lot of research, interesting research to do, and we have a lot of things to
29:07learn.
29:07And as long as we learn, I'm hopeful.
29:10But it's true, there are great challenges.
29:12And I also think it's a conversation we need to have.
29:18And it's hard to, when you wander in the alleys of VivaTech, it's maybe not the best place to have
29:26this conversation.
29:27But I really think it's important to emphasize that.
29:30What do we want to build?
29:31What world do we want to live in?
29:35The computers, they just compute.
29:37I mean, a lot of people in this, in VivaTech will tell you they reason and they think and whatever.
29:42It's not true, trust me, they just compute.
29:48The right question is what do we want to compute?
29:51I want to believe that we will want to compute funny stuff, exciting stuff.
29:57And it will always come from humans, artists, creators, engineers.
30:01I don't really make a difference between them anyway.
30:04But, yeah, I try to be optimistic.
30:08You, Anthony, optimistic?
30:10You know, I flew here from Houston.
30:12And I marvel at air travel.
30:16Countries all over the world have created a system making flying in a piece of metal 40,000 feet in
30:23the air the safest way to get anywhere.
30:26And connecting the world and countries that don't like each other or get along all agree to the same rules.
30:33It's, I think, one of the greatest triumphs of the human collective.
30:37And I wish the same thing for AI.
30:39If we can do it for air travel, let's do it for AI so it really serves humanity and makes
30:45it safe and wonderful.
30:46You know, air travel enables us to move our bodies.
30:51AI enables our minds to travel all over the place.
30:55Let's make that as safe and productive and wholesome for humanity.
30:59And, Nathalie, you are also optimistic.
31:02No, I'm not.
31:05Of course.
31:06Always.
31:07Yeah.
31:09We deeply believe in the powers of sound and music.
31:16It's so important, so beautiful.
31:19We need that to, not just to live, to be alive.
31:24So, it's, yes, as I said, as a neutral technological company, we think that we are useful not only for
31:35the music industry and the artists,
31:37but also for society to build, as I said, a model where new barriers, new frontiers can be built to
31:49prevent from fraud, from pirates, from dishonest behaviors.
31:57And, on the other hand, to build a new system where the remuneration for artists, for scientists also, I'm a
32:10scientist, so, for scientists also, put their work ahead of the, like, the mer content.
32:18It's, it's in our end right now to build those systems.
32:23So, that's great.
32:24That's a huge opportunity.
32:26It's not too late.
32:29Manuel, what are you working on regarding AI?
32:33What will be the next applications you will release?
32:37Or maybe, I heard you just opened a new center recently, too?
32:43Well, yeah, we try to extend the kind of research we do by working with social sciences to understand better
32:51why people listen to music,
32:53and running interviews and mixing methods from data science with social sciences.
33:00It's actually fascinating.
33:02On the more technical side, we work a lot on trying to investigate these generative models to see how, if
33:10we take them part by part,
33:12we can find ways to, I don't know, maybe identify in the future what were the training pieces in the
33:21training data sets.
33:22If we can identify that a model was trained on the work of art, and we can prove it mathematically.
33:30So, that's also the kind of stuff we try to do.
33:33But mostly, lately, I try to talk to people and convince them to work together.
33:39That's mostly my job, to talk to artists, labels, tell them all about what's happening on Deezer, on the streaming
33:48platforms,
33:48what we see, what's coming, and why we need to work together, actually.
33:52And we don't have all the answers, and there are stuff that I can't do.
33:56So, yeah, I try to make friends, I would say.
34:00Okay, it's always a matter of humans, at the end of the day.
34:04At the end.
34:05Well, so, thank you very much, the three of you, for being here with us to talk about this great
34:11topic.
34:12Thank you.
34:13Thank you.
34:13Thank you.
Commentaires