The Europe In Synch Podcast
The Europe In Synch Podcast
EP13: Act In Synch 2024, Athens Special - The Copyright Talk with Hans-Peter Roth, Alkistis Tsiklou & Martin Nedved.
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Join us in Athens, Greece, where the Act in Synch Summit is fostering cross-industry collaborations and sparking discussions about climate change, sustainability, cultural impacts, and urban transformation. The summit's vibrant setting at the Impact Hub provides an ideal backdrop for exploring how music intersects and acts with advertising, urban development, tech, and global challenges. We highlight insights from influential speakers & delegates who emphasize the importance of community and collaboration in overcoming industry obstacles and forging a positive future in a post-pandemic, AI-driven world.
The explosive growth of AI is reshaping copyright as we know it and this fascinating panel discussion brings together three industry heavyweights to untangle the complex web of music rights in the age of artificial intelligence.
Hans-Peter Roth of Muserk, whose company manages a staggering 15 million copyrights, offers a pragmatic perspective from the frontlines of rights administration. Alkistis Tsiklou of Musou Music Group brings deep expertise from a career spanning major labels to copyright management, while Martin Nedved of AIMS API contributes legal insights as an AI music search pioneer.
The conversation identifies a fundamental challenge: while the EU tries to move with unprecedented speed to regulate AI through its landmark AI Act, technology continues to outpace legislation. As Hans-Peter puts it, "This is an unfair arms race" where creators are struggling to keep up with rapid technological change.
What makes generative AI particularly threatening isn't just its use of copyrighted works, but how it creates new content that directly competes with human creators in the marketplace. The panel explores thorny questions about fair compensation models; how to track which works influenced a specific AI output; and ethical boundaries when it comes to artists consenting to have their work recreated.
The experts also raise alarm about collection societies being overwhelmed by millions of AI-generated works flooding registration systems although, despite all these scary challenges, there is cautious optimism that AI could eventually become a "force for good" if properly regulated and licensed.
The discussion is expertly moderated by Europe In Synch's, Hannes Tschürtz.
Europe In Synch is created, managed, promoted, and driven by several European organizations and companies and is a truly cross-border collaboration.
The goals are to bring together professionals from the music sector with decision-makers from film & advertising to provide a real-life, hands-on, learning experience, and to promote European music in the complex field of synchronization, through communication, knowledge-building and networking via focused mentoring and peer training sessions.
Follow Europe In Synch:
- Home: https://europeinsynch.net
- Facebook: https://www.facebook.com/EuropeinSynch
- Twitter: https://twitter.com/europeinsynch
- Instagram: https://www.instagram.com/europeinsynch
- LinkedIn: https://www.linkedin.com/company/europeinsynch
Intro/Outro music is an instrumental edit of "Gimme" by Daffodils.
They're on Soundcloud.
Europe In Synch is co-funded by the European Commission.
This podcast is a SuperSwell production.
Introduction to Copyright Panel
Speaker 2Welcome to the Europe in Sync podcast . We have another special episode coming from the 2024 Act in Sync Summit in Athens , greece . We're bringing you a selection of talks and panels featuring excellent discussions around the subjects of music , tech , brands and time . This episode is the copyright talk , involving Hans-Peter Roth of Music , alkistis Tsiklou of Muso Music Group , martin Nedved from Ames API , and all moderated by Europe Insync's own Hannes Tschurz On the panel . They talk about what the trends are in collecting copyrights , especially around the explosive number of registered works , largely due to the sophisticated and rapid advancement of AI systems . They also discuss what strict rules and guidelines might be needed to control this use of AI and , if it's all too late , get the AI genie back in its bottle . So let's dive right in and join Hannes talking to Alkistis , hans-peter and Martin .
Speaker 3Thank you for coming back . First of all , it was an exciting half day . Already . We have one and a half more exciting days ahead of us . Thanks for sticking with us . Thanks also to everyone who's watching online . Hi there , I feel the urge to say hello to you too . There's a lot of people watching online , actually , so we spoke a lot about artificial intelligence in the morning already , and we continue on that path . We realized it's a very , very , very broad topic . So let's see where we end up . About the copyright talk , I'd like to introduce the guests , or the guests introduce themselves . Let's start with Hans-Peter . Who are you ? What do you do ?
Speaker 4My name is Hans-Peter . When I'm not doing this , I either referee American football or fly glider planes . That's what really fuels me as well , but other than that , I do music copyright administration . My company , music , is the biggest rights administrator in the US market . Is the biggest rights administrator in the US market . Started with 30,000 works in 2017 . And , as of this summer , we now manage 15 million different copyrights in the US . We use AI for , of course , matching , but also for reconciliation of payments . A lot of
Legislation vs Technology Speed
Speaker 4post-usage matching is what we do . Try and make the money coming out with data from the users , the DSPs , make sense to go back to our customers in a seamless way . So it's very much the gray eminence part of the business that I delve in . I have not met an artist ever other than at a concert . So , no , I'm running the numbers , I make them look flashy , yeah , Martin , we've met you in the morning .
Speaker 3Still for the people who are with us .
Speaker 5So I'm Martin . I run a company called AIMS . It's an AI music search company . My background is in law . I actually did copyright law for a couple of years . I'm quite active in the copyright administration , cmos , etc . I sit on a board of a neighboring rights society in Czech Republic . I'm also chairing International Production Music Association . Yeah , looking forward for the discussion .
Speaker 3Fantastic . Thank you and Alkistis , do I pronounce it ?
Speaker 6properly . Yes , alkistis , yes , that's correct . So I'm currently Copyright and Royalties Director in Mosum Music Group . It's a 360 music company . I have been in the music industry all my working life , from Virgin Records to Warner , chapp chapel and now to muslim music group .
Speaker 3I turned from the recording to the copyright and never went back , actually , and I'm looking forward to our conversation when I lecture music business at universities and stuff , I always tend to say that really every little sense , everything that money in the music refers back to copyright .
Speaker 3And here we are at pretty much a crossroads of maybe even copyright itself or the music industry , with all the AI topics that we've discussed in the morning and also the backdrop to that discussion might be . If we look at legislation as it is , say , in Europe as a standard is usually lagging behind years , if not decades sometimes . I remember a particular case in austria where I'm from , about face button up . So the private copy , uh , so it came way too late , and when it was finally settled whether you should pay your extra cents on hard disks , basically the whole thing with copying music onto hard disks was over by then it it was like 10 , 12 , 15 years too late to actually solve that . What's your opinion ? Where do we stand with the whole AI topic in combination with where legislation and also the consciousness of politics is in that sense ?
Speaker 4I mean , this is an unfair arms race , basically because technology can move so much faster than any legal document can be created , even though maybe they should use a autogenerator is an unfair arms race , basically because technology can move so much faster than any legal document can be created , even though maybe they should use a autogenerator . That's another question . So it's just unfair . And it moves so fast and everybody's trying to get their head around it , because that's kind of the job of the legislators . They cannot just , you know , shoot in random places and then hope to hit the bullseye . They really have to take a diligent approach and we're having to do that with a human capacity mind compared to what's going on in technology , which is exponential . And this is not just an AI question , it's a general question . It is trying to keep up . It is almost an exercise in futility . So we are way behind .
Speaker 4But with that in mind , if you want to get paid , if you want to get paid , it doesn't matter what technology uses your rights . You still need to do the freaking homework and it's boring and it's tedious , and as long as it's the youngest intern that has to do all the registrations between getting coffee and food and cleaning the bathroom . You're never going to get paid . If anything , do your homework , Put down on paper who did what and what they get from it , so you can register it . Otherwise you're going to end up in a void and be frustrated and you're never going to get paid . So , if anything , technology has made the classic virtues even more important .
Speaker 5I can just go maybe into . Like , generally , law is always quite behind , but I would say in this time , eu is actually staying on top of things quite well , EU intentionally wanted to be ahead and they wanted to be kind of a role model for the rest of the world . So they did this AI Act and I would say we've only started talking about CheGPT , what , two years ago , I think , it literally came out in November 2022 . I mean two years . You know , that's like lightning fast when it comes to legislation process . So the fact that that has actually passed through , you know , european Parliament is quite surprising , and they have been very intentional about wanting to set the precedent for the other countries around the world and that legislation actually seems to be quite , you know , well thought through and it has had a lot of input from the creative industries music especially , like you know , the organization like ICMP , who spend like so much time on actually making sure that our voice is heard in there , and there are some really good things in that legislation . So it's now about how the individual countries also implement parts of it . Not 100% sure how it works , even though I'm a lawyer as well , but I haven't been doing it for 10 years .
Speaker 5But some of the things that are in the legislation are super interesting . It's saying very openly you have to get a license to train For this room . That's not shocking , but you would be surprised that that's not the , let's say , majority opinion everywhere in the world and it's not necessarily the majority opinion of the tech companies especially . So that's one thing it very clearly says you have to get a license to use the music for , or any kind of copyrighted material to use it for training . That's one thing . What is really important in the legislation is that it's also laying out it doesn't matter where you do the training . You can't do what is called forum shopping that you will look for the country with the most lenient laws to do your training in .
Speaker 5This legislation is saying if these models are being used in the european union , they have to adhere to the law here , which is really
EU's AI Act Leadership
Speaker 5important because eu is the second biggest market in the world after the us . So if you want to use your AI model in EU , you will have to adhere to this . You have to keep track of what are you training on . That's super important , so you can't just hide behind .
Speaker 5Well , we used whatever we scraped somewhere , which quite often is the answer of some of the tech companies . It's like oh , we don't even know what's in the model , because you know we scraped whatever is publicly available . That the model because you know we scraped whatever is publicly available that's what they , what they quite often say . Well , this legislation is saying you will have to keep track of what is actually in your data set . So it's starting point and I think it's it's very good . Now the question is how exactly it will be used , how exactly it will play out and how it will play out in the other countries , in in us , especially , and in us . I wouldn't bet that that's how they are going to implement it .
Speaker 3Okay , but Akisis .
Speaker 6Okay , I'm not a lawyer , so I will say things how we see it from the copyright side , and we face right now a problem which is AI . We need legislation to make sure that AI is properly used . We need time because the implementations need to be done in different countries with different laws , with different approaches . Some Asian countries still believe that they need to have an open source for AI and they can ingest anything without any regulation , so there are still many things that lawyers need to resolve . From our part , we want to make sure that anything that goes in is either used without any commercial engagement to follow or , if it is with a commercial engagement , then needs to have received a license , the writer's consent , and the writers have received
Licensing and Training Requirements
Speaker 6their compensation .
Speaker 3I wonder how this happened again and again and again . If you think about how YouTube , for instance , was created , basically Pandora's Box was wide open already when someone realized , hey look , there's something that actually belongs to me and I don't get compensation . And it took years and lawsuits and so forth a bit the same problem , as we figured with legislation in general . But the compensation , on the other hand , especially for those who create the works , is one crucial point . You might realize that just last week , the AItrainingstatementorg website went on and tens of thousands of artists and organizations signed this very simple one sentence that I'm going to read to you now the unlicensed use of creative works for training generative artificial intelligence is a major , unjust threat to the livelihoods of the people behind those works and must not be permitted . But all these platforms are already using it . So how do you get that back in the box ? And is there a way to even get it back in the box , to get to close Pandora's box again ? Martin ?
Speaker 5Well , I don't know . I mean , like you have probably all heard that the two companies Suno and Udeo are being sued by pretty much everyone , like in the US . They're the ones who admitted or at least one of them did that they trained on everything and their position is well , that's fair use in the US , because they're trying to claim that it's similar to how musicians , artists , to claim that it's similar to how musicians , artists , composers train on other people's music as well , basically having the theory everything in the internet is there and it's available , so it's free and I can use it yeah , yeah , that's pretty much their theory .
Speaker 5And I mean , like , for us I think it sounds shocking , but I wouldn't completely rule out that a court in the us might rule in their favor . I'm not saying it's probable , I'm not saying saying that's like you know , it's 60-70% , but like you know some of the rulings in the past , I wouldn't completely rule that out in the US . In the EU I would probably rule it out , but in the US I don't know . Let's say , 20 years ago , if you told someone , okay , google is going to scan every book in the world and they're going to make it available for search and to read parts of it without license , would you think that that would be okay or not ? And then the court in the US said that's totally fine , that's fair use .
Speaker 5And I'm not saying that I agree or disagree with that , you know like that's not my position , but like that was probably as shocking to the book publishers as it would be for all of us if the court in the US said , well , if the court in the US said , well , no , that's actually fair use , I think everyone is kind of silently counting on that 100% happening that like , oh yeah , the Suno and Udo , they are going to lose . I think they are going to lose , I think so , but is there like a part of me that is thinking , well , maybe they are going to win that , and the reality is it couldn't happen ?
Speaker 3I mean , the one sentence sounds incredibly vague in the first place . It can be really everything , but it leaves out a couple of really critical questions , which also reflects on your work in particular . I mean transparency , the sheer flood of data that's involved in that . How could you possibly say what AI has been trained on and how to walk that road towards an actual , fair compensation , and what can that compensation be ? At the end , it just states , okay , give me money , but if we think about the ongoing discussion about what Spotify pays out for a stream and this training model obviously can only be a fraction of even that so really , from the technical point of view , almost how could this work ?
Speaker 4at the end . I mean , being commercial in this space is dead easy . You just come in with a number and then you hope somebody's going to pay it . So according to Gamer , it's 30% that's what they say 30% of the revenue generated by the platform that has used the music . Now here comes the interesting part . Is it about how much of the data is music and how much is something else ? Is it the byte size of the actual data ? Is it the usage in the model ? Or you might have something available , but it's only been used a very little , scarce part of the time .
Speaker 4So here it becomes very interesting how you want to calculate this , if at all . So the brute force is saying it's a blank tape , maybe 30% . Just give it to us and we'll algorithmically , in an analogy , will , distribute it to writers like that . But if you need to become more surgical , it is going to be a very interesting one , because what we also know is all these platforms say that by divulging what data I have on my platform and how I use it , I am also divulging a business secret , which is the secret sauce on how I then generate this commercially viable output . So there is a tug of war that is going to happen , but , yeah , 30% according to Gamer .
Speaker 5And I have a question to that . Would we be okay if Spotify was paying 30% to the artists and composers ?
Speaker 4That's a huge increase .
Speaker 5Well , now they are paying 70% , right .
Speaker 4Gamer is only composers and authors . This is not the sound recording part of things .
Speaker 5But aren't they saying that for all ? Are they saying only for the composers ?
Speaker 4That's what Gamer is saying 30% .
Speaker 3Creators of the work , meaning composers and lyricists , yeah .
Speaker 5Because some other initiatives I've seen that we're only asking for 20 or 30 , but over the total for everyone . That's why I'm kind of asking these what I would call like aggregation plays . It works usually for the platform . It doesn't necessarily work with the artists or people who are being aggregated , because let's say that the sunio and udo will make their whatever . Let's say that they make a billion and then they distribute . Okay , let's say that 70 percent , like spotify duty , whatever . It's going to be 100 million people that were involved in that , so every one of them is going to get 70 cents .
Speaker 6If I may add something , there is a difference between platforms using copyrighted works and how fair it is for the writers to receive and the right holders , their shares , and there is a huge difference between generative AI that will , by using and being trained with copyrighted works , will finally create new works that will end up to compete with the copyrighted works in the real market .
Speaker 6So the difference is the commercial engagement of it . Generative AI will not be a platform that copyrighted works will be on . I may take 30% , I may take 40% , Whatever is best for the right holders . The problem here is that you trade something with copyrighted works and the result , the outcome , the commercial use of it will actually compete with the original works and with the human writers and right holders . Because YouTube , yes , was a platform , but when YouTube started to commercially use and become the music business by itself , then it needed to receive clearances , consent , licensings and pay the right holders the right money . This is the problem and this is all . The writers fear that by using their works , they will create a result that will finally compete with themselves .
Speaker 4And if I can just scare the little impetuous out of everybody in the room , I think is how many iterations before something disappears ? How many iterations of AI , algorithmic use of the original songs before their individuality disappears ? So to your point , for instance , what if I make an AI service that generates songs free of charge ? Now I'm outside of the commercial loop . Now you just ask my service , free of charge , to generate 1 million songs and you put them into the commercial model . Is there a link
Generative AI vs Platform Models
Speaker 4or is it something new ?
Speaker 4So there's tons of things here that's going to be blowing people's mind . It's going to make it really , really scary and at some point of time , I would say that we're getting scaringly close to getting inspiration . As you said , at some point of time . Everything has been there already . That's 12 notes . I mean , how far down that rabbit hole can we go ? So it's a very , very weird situation and I think that a lot of the work that companies like mine have to do is to be very good at helping going back in time and claiming for past transgression . That's going to be the biggest one , because going forward it's going to be a different dialogue , but there's past transgressions .
Speaker 3That needs to be paid for for sure , and that's where data capacity is going to be hugely important it comes back to the transparency and data point , and also I think Paul McCartney said in 1969 or something like that all the songs have been written already , there's no point in continuing , and that's quite long ago and the number of songs increased quite drastically since then . So maybe AI can foster inspiration in a way too , but if we bring it back on those tracks it's somehow a gigantic brain fuck really . I mean , we we're discussing basically whether for collecting societies this is an individual or collective right . We don't have a clue how to actually show which songs ai has been trained on in a really transparent way I would modify that the ai companies do know what they train on .
Speaker 5They know what goes into the data set . That's not really the problem .
Speaker 3I mean , of course , you can be A specific song or as a model .
Speaker 5You can be negligent , like , not keep track , but you can keep track of . These are the files that I'm literally putting into the training . The really difficult part is when I spit out a new song , show me the tracks that went into producing that specific song . That out a new song , show me the tracks that went into producing that specific song . Because , yeah , because that's like right now , as like when I'm reading really smart people they are saying that's not really solved at this , at this moment , because , like , really , what went into training that one song are all the songs that you put into the input because , like , even the ones who are really bad examples of that , you use them as a negative example and you use that in training actually quite a lot . You're not just telling it okay , I want a rock song , but you are showing it and this is not a rock song and this is also not a rock song and that actually improves the quality of the output .
Speaker 5So , should you merit only the five closest tracks to the one that you have spit out ? Because that is doable . Like , we have a similarity search so I could tell you okay , out of the data set that went into the training , these five tracks are most similar , but that's not really the thing that we are asking for , right ? Because maybe these five tracks were not really that important , maybe they are mediocre tracks . So the problem is really I don't even know whether that's solvable , like and I've read quite a lot about that , and I don't even know whether that is really solvable to say that these are the 10 tracks that are most important for training of this . The first part is super easy , like what went into the training of the model .
Speaker 4Yeah you if you tell , if you report how you use the music that's in your database , you're also telling what the negative input value is and what the positive input value is . And now you're starting to divulge how your entire machine is working , and now that can be copied . So , funny enough , people who don't believe in copyright don't want to have their stuff being copied . Go figure . But that's part of the data problem , is ? It's going to be very , very revealing .
Speaker 3Plus , you have extra layers when we think about copyright and how it is created right now on both ends . The recording side , with its 70 years , and the actual copyright side , depending on the lifetime of writers and lyricists and composers , can be quite confusing , so it's a huge mess . Basically , I think it would be a nice art project . There's been I always forget the name of the artist some 10 years ago , an artist created with one of the very first AI models , all the possible combinations that there are to melodies and rhythms that are known . So it's pretty easy by now . It's basically at the push of a button you can create all possible copyrighted works , and he tried to register that with Gamer , basically proving the point that all the music is there already .
Speaker 3So anything that wants to be copyrighted right now does belong to me already , but then giving it back to public domain . So music is free . Finally , it was an art project , so it didn't really turn out the way he
Transparency and Data Problems
Speaker 3liked it to be , but as a principle of thought that we follow here too , it's practically impossible . So it feels that we are near the end of the road of creativity for one , near the end of the road of what's bearable for computers in terms of well work with data . Where do we actually go from here ? If you look , say , five or 10 years forward , where would your companies be and what they've done and achieved with that topic that we're talking about ?
Speaker 6I think that the first thing that we must say is that as better as a work is labeled with all the common ways we know right now with ISWC , ISRC , with the writer-composers , writers , companies , record companies the better chances we'll have in the future to be recognized through generative AI and the outcome . Tech companies believe that there is an imprint there , as for any other movement in the world of the internet . So the words that are correctly labeled , they will have a chance to survive this mechanism and find themselves copyrighted in the end when we are there . So make sure that all your words are labeled correctly from the record companies , from the publishers , the copyright societies , from everywhere . It has a chance to survive .
Speaker 6On the other hand , there are discussions about ethical and unethical generative AI companies . I don't know the example you mentioned , but we have the example of Andy Travis that lost his voice due to his stroke and by turning to the so-called ethical generative AI companies , along with Warner , along with his producer and along with his consent for the whole project , he created a song with his voice from previous songs that entered the charts . So could this be considered as a good example of a generative AI ? Could this be considered a copyrighted work , since it has all the proper consents , Because it seems that the works coming out from generative AI won't be copyrighted , as everything looks right now .
Speaker 3Which leads to the excellent point of where you draw the line between ethical and unethical . So we recently discussed a very similar example . So this would be ethical because of his will and consent to use this , to use his own voice , for that very purpose . So imagine an example of Leonard Cohen singing a song from Whitney Houston in Greek language . It would be possible , it's quite easy to do so , actually .
Speaker 6But we don't have his consent .
Speaker 5Well , his heirs might give the consent , or the record companies who own certain parts of the recordings . They will probably not own his moral rights . These will probably pass to the heirs and I would say that if the heirs give the consent , that's probably as close as ethical as you can you can get which leads me to exactly the point where I wanted to be , thank you .
Speaker 3so the moral right is not part of copyright at all , and I mean only between the lines , if you like . Is this something that the industry is going to exploit in the future ?
Speaker 5I'm not sure if I should be answering that . I don't know . You want to answer that ?
Speaker 4I'm just the great eminence , I just do
Future of Music Copyright
Speaker 4whatever people tell me to . So I'm pretty sure that , when you look at the challenges that , for instance , cmos puros have in their going forward , I think that the moral rights of things will be a very , very high priority in their existence , in order to make sure that these things are not disappearing . But I don't think it's going to disappear because of of this alone , and I would hope not if I can go back to what you asked before about this is the end of the creativity .
Speaker 5I definitely don't think so . I actually think that this might be a force for good if used in the correct way , and I actually think that we are probably going to get there . That it might be very messy and we have to work for it as well , but I think if it's properly licensed and then let's leave now what the definition of properly licensed is , and that's for , like better lawyers , to figure out what that , what that means and I think it can definitely be like force for good . You were discussing earlier that , like how amazing it is that now any artist can take their laptop and create from anywhere in the world , and if someone from the 50s heard that , that would be a travesty for them . Like , what about all the session musicians ? What about all the bands that lost their work because now the composers and the songwriters they don't need them .
Speaker 5They are just using their little computers to do something that's imitating what the bands and the musicians were doing before , and I'm pretty sure that that argument has been used many times , right , like with the vinyls and with first recordings . Well , the musicians were striking against that .
Speaker 5They were saying this will kill the live music with every technological shift , I guess exactly , and I don't want to underplay that I'm not necessarily saying that this one is exactly the same , because I think we all kind of feel that the ai is slightly different , that it might not be exactly the same like any technology like before , but at the same time we should acknowledge that , like already , where we are today would feel like a yeah , travesty to people who were in the music business , maybe you know , 60 , 70 years ago so we can even go further back and go .
Speaker 4Well , where does music lie ? Does it come out of suffering or does it thrive in the seamlessness of things and maybe of all these curveballs that consistently has been thrown at this industry and other industries as well ? Maybe that's part of fueling the creative process . Maybe that's part of trying to tell a story . Maybe that's part of what you're all about here in Acting Sync is , how do we create content that does more than just pass time ? How do we create content that can convey a message that has deeper meaning ? I just thought it was very interesting that maybe suffering is part of the DNA of this industry .
Speaker 3Not just the industry . I would go even further . Music has always been an agent of change and this is one of the very topics of this conference too from many different perspectives , and music was often really on the forefront of technological developments , evolvements in societal views , many different perspectives from that . And once again , it's not only music in that case . But when it comes to the very questions we discussed right here , we don't really exactly know where this leads us . I'm thinking of the Grimes examples . That's now some two years old already .
Speaker 3Grimes , an artist from the US , allows people to use her voice if they agree to share the profit by 50% , I believe . So already in the US they're making business models out of what is a problem for other people , and we're still about to define actually the field , how and what and why in this very field . How long would you think does it take we finally end up in a rather comfortable place where we at least have the ballpark numbers or something we have with the likes of Spotify and Apple right now ? How much time would you give yourself to create a service , a meaningful model for that ?
Speaker 4Give me 30% and I can analogy distribute that like a mofo . That's not a problem . So no , that's the easy part . You can always get money out of users . It's more a question . I've always said I've been in the front office part of this business doing the deals for the X percent and then I say I move my way up into the middle and back office part of the industry , because how do you get the right money out to the right people ? That's much more interesting . So you know fleecing in lack of a better word companies for using music . It's not difficult . They write a check and they do a bank transfer . It's more about how do we get it out , and that's much more problematic .
Speaker 4Here , I would say , because of the nature of the data and its usage . You could be rudimentary and use a very basic analogy , or you can try and get very transactional and then that's going to be a huge problem . Then I just want to say another thing .
Speaker 4Here is the biggest problem for collective rights societies and PROs right now is not so much this use . It's actually that right now people are spitting out tons of content and are registering it with CMOs around the world , and just because it's out there doesn't mean it's going to get used , right ? I mean , availability is not the same thing as consumption . So all these PROs are picking up the tab on behalf of the creators to register all these works that are never going to go anywhere . They just lie there dormant . You saw that only 15% of all available tracks has over a thousand streams and 25 has nothing , right ? So I mean all this dormant stuff . Somebody has to pick up the data bill for all that management and for the correction of it . That's probably the biggest threat right now is you're paying . Creators are paying for nonsensical operations right now to an extent I don't think people fully appreciate .
Speaker 3I'd like to open the floor to you now . If there are
Ethical vs Unethical AI Use
Speaker 3any questions or something that's on your mind , you're very welcome to contribute .
Speaker 7Yeah , marcel Alexander Wiewenga , ceo of the firm . In relation to what you said about , the problem is that it's replacing something , right ? That's where it becomes a trouble . So an AI gets trained on copyrighted material and then it creates something that replaces it . I think that's very true . That's where the problem is . However , how big is the problem ? If the quick calculations meant everybody was getting 70 cents , I think actually it was 7 euros , but one you miscalculated , but the value is just very small in this mathematical example , so maybe we're talking about nothing .
Speaker 7Secondly , to add to that , if it's about replacement , replacement is the business of every production library music that is out there . What goes for a computer system should go for humans doing this as well . You cannot let the humans go out of the legal loop , is my humble opinion , and as a person that used to run a company that thrived on this . Basically get three reference tracks in the training data , then have a prompt , the briefing , and recreate a track that was actually replacing an original composition . There are only in the Netherlands already 50 companies that do this on a day-to-day basis . I think in Europe there's around 2,000 of them that do this on a day-to-day business . The problem is not just about what computers can do , what AI can do . I think it's about how we look at the value of music , what is original and what is replacement . That was my thought .
Speaker 5I'll have to say something to the production music part because , of course , having my production music hat on , I think this is like a really big misconception about production music that , like a majority of the source of the income is sound likes , reversioning , et cetera . It's for most companies it's a marginal source of income . We have done studies in several countries 12 countries actually , and in several of them I would say seven there are Piero blankets that allow the broadcasters to use whatever music they want for their own productions . So you have a Piero , you have a broadcaster let's say RTL in Germany and they can use whatever music they want commercial production music , whatever they want . And they still overwhelmingly use production music , even in those countries .
Speaker 5So it's not a . It's not about that they can't use the , the original that we would call . It's not about the price either . They pay the same . They already paid the blanket . It's actually that production music is created with . The purpose of . This is going to go into the audio visual . It's better suited for most of those types of usages . So this is like really like one misconception that I find very , very often that , yeah , production music is used because it's easy to license . That's true , but it's like I can't get Madonna , so I'm going to get something like Madonna from production music . It's not that they can't use the original , they just don't want to for some reason . For production music , really , the gist of the income comes from broadcasters , from Netflix as well . For the companies that I've been involved , netflix is a huge source of income as well . They do have a blanket for that as well , not for the sync , but they do have for the performance .
Speaker 7From my own experience , what my company used to do is replicate copyright , the same as AI is doing right now , and I think the solution is maybe more comparable to what sampling has been doing . Sampling at first came out also technological revolution and they didn't have any way to monetize that , but obviously it should , because the songs are being built on the back of other copyright and now the same thing is happening out here . The big challenge then is if you generate something , what is actually being sampled and then do you do a lump sum and give everybody their 70 cents or seven euros , depending on the calculations or can you sort of retract okay , this song is generated under these hundred tracks by the Ramones and therefore you now have a Ramones sounding type of song that you're using .
Speaker 3That's going to be a very interesting question for the future . We have one last question , nies .
Speaker 8Well , we're talking about creating music that is a replicate of others' I think for many years the radio format has squeezed music production into sounding equally the same . You know , we have tons of ed sharon's versions out there , so haven't we been doing that ourselves for such a long time ? So I actually think that this time is more inspiring the people to sound unique and interesting , maybe leaving a space for creating your own voice and your own sound . But the ones that still follow the trend and wants to be a radio hit sounds more or less the same as whatever is played on the radio daily . So if we are pointing fingers at AI or whatever , I think we put ourselves in the situations copying ourselves so many times . So who owns what ?
Speaker 3Actually an excellent point to bring in some useless knowledge . The University
Questions and Final Thoughts
Speaker 3of Vienna did a study two or three years ago proving exactly your point that music on the radio is sounding more and more the same . And I mean this is a very , very different discussion , but it brings me to my end note because we're running out of time . I'm a fan of graphic novels . I don't know if you are , but if you go onto your favorite browser there's a search bar and you enter theft , a history of copyright . It's a fantastic graphic novel that's highly entertaining 300 pages . It's for free , you don't need to steal it . It's really very , very entertaining . Actually proving or building the bridge between what you just said , what we've been discussing , maybe showing us that , after all , it's not that bad to borrow , steal be , inspired by Was that the outcome of the panel ?
Speaker 5That's the outcome of the panel . Did we say that I ? Feel abused in that .
Speaker 3AI model . But thanks so much for your contributions . Hans-peter Martin and Alkistis , thank you . Thank you .
