Hello and welcome to the Sound On Sound recording and mixing podcast with me, Sam Inglis. Today I'm joined by the CEO of Sonarworks, Helmuts Bems. Delighted to meet you. Nice to meet you, Sam. Thanks for having me. Thank you so much for joining us. We're talking today about a survey that Sonarworks has commissioned in partnership with Sound on Sound about musicians and producers attitudes towards artificial intelligence in music. Sonarworks has always been a research based company, but most of the research I've seen from you in the past has been around room acoustics and headphones. What made you decide to commission a survey on this topic? Well, first off, I'd like to take a step back in terms of that last comment that you made. We indeed are a very research based organization. We do, technology wise, we do a lot of, we call deep tech research. So there is a fundamental sort of technology aspect that we are working on. Some of it is gut engaged, that's never been done anywhere else. So there's that part. But as a company, we are extremely data driven, and the core philosophy internally for us is to make decisions based on data. When you work in the field of audio, most often you come across, you know, goldener concepts where somebody you know, with a label of golden ears ultimately gets to decide what is good or bad. And there's nothing wrong with it per se, but as our company principle, we. Very early on decided that we will, as much as possible, try to base our decisions business wise or technology wise on data. So a big part of what we do, we try to kinda make the decisions based on data, which means we have to gather data in the first place about all aspects, and that goes technology as well as business. So I think relative to other companies. We do gather quite some more data for decision making and that also affects the business side of things. So we do a lot of research on market users, their habits and how they feel about things. So that's an ongoing thing. And then slightly more than a year back, when we were thinking of starting to go into AI tools, uh, for music production, it naturally started with a lot of research into the field. And last year we published a white paper on the state of AI for the music making. And that was sort of qualitative research. So we kind of went into having, uh, lots of interviews with people going in the details of how they feel, but we were trying to kind of get the pictures of how the field looks at that moment. So we felt at that point, kind of qualitative research going deeper with fewer people made more sense. And now, we launched the product voice ai more than a year ago, I think a year and a half. And now we felt like it's a good time to look at the topic quantitatively, like continuing that research that we had in the field and then we and then kind of we faced the challenge of how do we actually scale it up in terms of data gathering. So sound and sound, that felt like a perfect partner because we wanted to get away from the bias that we might have. If we would just conduct the research ourself, because naturally we target specific audiences and specific people that are within our reach. And then we felt like that would be biasing us in terms of the conclusion that we made. So Sound On Sound felt like a really good partner in doing this. And then, uh, yes, we scaled it up and kind of in a nutshell, we got to 1,200 users that were surveyed. The most important thing for us was to, yeah, avoid the bias and then kind of have as general and wide reach as possible in order to get a good picture of where it is. The ultimate goal. I mean, for us it's a lot about business decision makings, steering, where to go with the product development and what the needs are for people, for our potential users, as well as kind of assessing, you know, technology aspects. Assess where there are gaps in terms of technology and what areas need more attention. So yeah, again, the basis was really to base our decisions on, on objective data rather than somebody's opinion. And when it comes to AI, naturally there are lots of opinions around, right? And people can get quite emotional about the topics, and that includes us as well. So our goal was to get to a more objective reality. Absolutely. People sometimes divide AI tools into two categories. There's assistive AI and generative ai. Assistive AI tools are ones that help human creators do their job faster or better, whereas generative AI tools are ones that, to some extent replace human creativity. Do you think that's a, a helpful distinction for understanding ai? Yes. Well, that's a very good question. This kind of division into these two groups was the one we came up with when we did our first research, uh, like two years ago. The qualitative one. And then, uh, this way of looking at AI products really came out of that kind of research. Research. I think it's still a really good way of looking at things. I do feel like borderline between these two use cases is somewhat getting more blurred at the moment. I believe it'll get more blurred going forward. And I think the one I, I'll give you a very specific example of where I think it kind of, where is exactly this middle line? We are not yet there. I think it's just a matter of time when we'll start getting products in that gray area in between. Let's say, let's first define the two extreme examples. One is, um, fully generative, uh, content, which, uh, my definition of that is those are tools that basically only have prompt as an input, which means the whole creative process gets done by ai, and AI just fully generates the end content. And then on the other hand, you'll still, you have AI assisted, which is people are still making the decisions and AI is just assisting humans as a tool, let's say some mastering tool or mixing tool or restoration tool that just, uh, maybe sometimes takes care of the technical aspects of the things. And to me, a good example of something in the middle would be. Let's say, imagine that you are just sitting down and playing your whatever instrument that you would be playing, let's say as a practice session, and you have AI to play along with you. Let's say as an example, you play guitar. Now you have AI accompanying you with a bass guitar, and that would be to me, an example. Exactly right there in the middle. It's still you kind of participating in the process, but it's also AI generating something fully, but then in line with you. But I think the division in these two categories is still a good one because we are technologically not there yet in terms of these products in the middle, but I'm sure that they will eventually emerge in just a matter of time. Absolutely. And one interesting thing I think we're witnessing with AI is the emergence of new categories of tools. For instance, there are AI focal tools that are on the one hand restoration tools in that they're designed to work with a compromised source, but they're also creative tools in the sense that they're re-synthesize the voice from scratch. They're not restoring the original recording. Do you think we're likely to see the emergence of more novel tools in this space? At the moment, I don't feel like I have, uh, I'm a good person to kind of predict what those are. I think those use cases will be kind of really new and I'm, I'm sure that they will emerge, but I wouldn't be able to kind of give you a list of what those are. If I would be able, we will probably be working on those tools. Yeah. Yeah, so back to the survey then. I mean, what we see from the results of the survey is, like you said, we had more than 1200 people participate, of whom the majority were professional, are professional musicians and producers. Um, about one fifth of those identify themselves as early adopters and about the same proportion are already reasonably heavy users of AI tools, but the largest proportion. Uh, would describe themselves as just beginning to experiment with a ai. Do you think we're at a sort of tipping point with the technology here? Yeah. Uh, and that was probably. One of the data points that I was mostly interested in, like internally we have had a lot of debate in terms of where the technology actually is in terms of production and I think what the survey now shows, let me just go through some of these key numbers. So basically out of the survey, the professionals, nearly half, 48% were saying, uh, that they are occasional experimenters. 20% said they are regular users. Another 13% said that they are planning to adopt, and only 18 said that they are not interested. Now I think the context, at least for me, is that. These tools appeared in the market, like really maybe two, three years. That's where I think, I mean, sure there were some tools that were early on, but I think the adoption really started, let's say it's been just two, three years into these new products and what it seems like is that. We have a solid 20% that has actually adopted the tools already. And that of course corresponds to early adopters, which is a, a product lifecycle kinda theory, right? Like, uh, most people have heard about it. There is a normal distribution. Usually that happens as new products get adapted. It's a bell curve. So initially it starts slower as early adopters are testing out the product. And then when they are finished testing and kind of have the first reviews, then there is early majority that starts using the product. And then by the time that half of the users have adopted, you have like late majority that starts, uh, looking at the product. And usually actually these product life cycles take quite a bit of time. It can be decades for this to play out Now. I think we've seen this in other general AI applications like with chat, uh, you know, AI tools where it seems that it's probably following the same adoption curve in terms of some people being early adopters and some people, uh, later adopters. But it seems to be squeezed in a much shorter period of time, and frankly, going through the 20% sort of early adopter stages. Within two, three years is incredibly fast, uh, by standards of other kind of products that's really, really fast. And this means that the technology itself is now at the stage where it has to start target early majority, you know, people that are not necessarily experimenters anymore in, in regards to products. And that means that these product offerings now have to change. These later adopters, they are because they, their use cases are not so much experimenting as, you know, finding a practical use for a product or a tool. So they will need much more kinda robust product offerings. And I think that this means that we will see a somewhat of a slowdown in terms of production because psychologically these, uh, majority kind of users. Majority in terms of, um, adoption cycle. They are more pragmatic and kind of slower to adopt things. So I think now we will face, like all the industry in general will face more skepticism. So there will be much more convincing, more robust product offering. In order get forward. Yes, and I think that the phase of like the super fast explosive user growth is kind of over. So we actually see a slow down in simply because so many people already have adopted. That's an interesting point because there's also a tension in the results here. Um, most people who responded to the survey I think can see the potential for AI tools to benefit them, but at the same time, there was quite a strong general perception of AI as negative or at best neutral. What do you think is responsible for that tension? Yes. And this, this is probably also like, uh, another super important aspect for me. The way I kind of look at it is, well, first off, let's take a step back to this earlier question that you asked in terms of generative tools versus like AI assisted tools. I, I think we have to be. Clear about like the audience and, and that's maybe another data point I quickly go through in terms of who were actually the respondents in this survey. So first of all, in terms of professional, by far the biggest majority were seasoned people in the industry. One third of those were over 30 years in the industry working professionally. And another half, another 50% are 10 to 30 years. So basically we have like almost 80% of the respondents are 10 or more years in the industry. So that's one. So that kind of means we are somewhat leaning towards more experience user base. And this is an important thing I think to note. And then in terms of occupations, they are 43%. They are mostly working as producers. We have another 20% of all the engineers. And then, uh, in terms of like larger groups, uh, 14% said they are, that they identified as songwriters. So, uh, this means we are targeting, uh, a somewhat more professional end of the spectrum. And, uh, newcomers were somewhat less represented in this, uh, survey. Now, if we look at the back of the division between generative and tools. These people that were, uh, like the respondents in the survey, they are mostly people that have already mastered their craft and they are mostly, and that was also coming up in the data for efficiency. They are mostly looking for tools that can solve sort of more technical problems that they have, and they don't really want the AI tools to kind of mess with their creative process. They would like to take care of the sort of dirty work, right. Some repetitive technical skills that they require. Now, I want to note here that, and, and this more falls into this AI as a helper tool. The other part, which is the generative tool, there are several aspects. First, I don't think that these particular, you know, producers and professionals are actually the objectively the right target audience to ask the question about regenerative AI tools. For one, because these tools are a direct threat to their profession. Uh, I mean, if generative ai, you know, completely takes over, ultimately, I mean, cuts out the creative process and this is what these people do. So there is naturally a bias there. Where they are leaning towards AI as assistive tool. And of course for us as a company making such tools, that makes perfect sense. However, generative tools I think are still a risk. And I think that I don't, I think the skepticism, uh, that people have is mostly coming out of the fear. The generative, you know, how the generative part of AI technology is gonna evolve and if it eventually takes over, I mean, it threatens that creative profession that we all love. And that's, I think, is the fundamental aspect of the skepticism. The other thing that really showed up in the survey was I think the ethics behind it. It was kind of very clear from the answers that, uh, at least the audience that we now, uh, gather data from is really, uh, really cares about the ethics of it. Uh, meaning that the models are trained, uh, legally and that the artists, uh, and the work that has gone into that training data. Has been, uh, you know, acknowledged and, uh, there is an ethical business arrangement behind it. So these, I think, are the aspects that create the skepticism. So from this perspective, I think that there is still a gap in terms of the ethics of it. So most people, I think overall, and I think probably if I look at the product offerings, there is still a gap in terms of how that is communicated in the legality and ethicality of these AI tools. So I think that's one driver and the other Yes, is, uh, an, I think somewhat objective fear from the generative part of ai. Even though, uh, the survey showed that the professionals that we were targeting, they are not really heavy users of generative ai, I was expecting somewhat that use those use cases to be bigger. But it seems like if I now, um, recall the data, only 20% of the users reported that they are using the generative tools. And, uh, that could be, that doesn't mean that they generate the full content out of it. It might be for ideation or, uh. For some creative idea generation process, however, it was only 20%. So I think, uh, there is a slight bias there. If we would look at other audiences, uh, we might see actually a bigger adoption of the generative part, but at least when it comes to professionals, they don't seem to be using the generative AI that much. At least other tools were actually more higher, like the usage was higher than, uh, for the generative uh, stuff. To return to the ethical issues for a minute, then do you think there's more that AI developers should be doing to A, to make their products ethical, but B, to communicate that they're doing that? Well, I, I think there are really two parts to it. First, that there is the communication part and there is the underlying substance part, right? Like. Not everybody, I think can actually communicate the ethicality of it because, but frankly, I don't think that everything is ethical from that point of view. There are definitely stuff out there that is abusing, uh, the rights of, uh, other artists before. And also we have to be careful when I say abusing, I have to be, I mean, I have to acknowledge that. We still have massive gaps in terms of the actual legal framework, right? Like most of this stuff actually falls in the gray area where there is no actual regulation, right? For how you should be. Like what is the authorship when it comes to training, uh, AI models and how the royalties work after that. So I think there is the actual gap in terms of the legal framework. And then of course, I think that. Even with that in mind, there are definitely tools out there that are pushing that, uh, boundary in terms of how it is used. And if you look historically of how technologies are getting developed, it's almost always the case at the beginning. You always have some kind of use. That naturally, uh, you know, ambitious business leaders will exploit, uh, in order to get some competitive advantage. So there's that part, but I do agree that there is more work to be done in order to communicate the ethicality of it. And I think, uh, I, after the survey, I also took it, uh. And discuss it with my team because I also think we have to do a better job in communicating, because in our case, because we were targeting a professional use case for the voice AI product, which is a vocal transformer, uh, it was very clear to us from the very beginning that like, we don't see the use case of, you know, replicating a famous artist voice. I mean, it completely doesn't make sense. Uh, when you are actually working with, uh, you know, reporting, uh, professionals, there's no use case for that. So we only use the voices that we have explicitly, uh, vocal models that we have explicitly, uh, contracted to be used for these AI models. But I don't think that we have done a full, uh, disclosure and always communicated it clearly. So I think there is a job to be done there. In terms of distinguishing, uh, you know, what is, uh, legal, because that's our advantage as I see, and I don't think that we have fully actually communicated the, that aspect. Do you think there's room for an industry-wide sort of certification scheme or something like that? Definitely. But I think, uh, I mean I don't think it's gonna come about anytime soon. Uh, it's a very complex topic because you can look at it from a point of view of the legal framework as well as just industry coming together and agreeing. And I think that will be very difficult because of very widespread interest the various parties have in this. So I'm sure it'll come eventually. I don't think it's gonna be a quick solution. So, uh, and I think it's beneficial. I think it's good to have one. They just, um, like, you know, being realistic, I don't think it's gonna be happening anytime soon. One of the really interesting questions in the survey for me was the one that asked producers about which skills they see are going to be important going into the future, and there seems to be a perception that what you might call power user skills, such as the ability to do detailed manual editing are going to be become less important, but that it's really important to maintain top level creative skills, musicality, critical listening, songwriting. There's also a strong emphasis on skills like networking, marketing, and the ability to work with other media such as video. My question here is, do you think this is specifically a response to ai or is, is the business going in that direction anyway? I think that when, I think in terms of where is the business going overall, and of course everybody whom we call professionals do care about where the business is going. That's what they're called professionals, right? They make that their living doing this. So I do think that where business is going mostly relies on way where AI is going. I mean. Frankly, that's the elephant in the room. Like even if you don't talk about it, I don't think it's realistic to look out five years into the future and not have a major AI aspect into the future. Right. At Sonar Works, we end the process of planning out for 26 and, uh, and beyond. We usually plan in a five year timeframe. And then, uh, for the last days I've been sitting and kind of trying to think of how the industry looks like in five years and there's just no way around it. I can't even fandom on, uh, a world where in five years AI is not playing a massively bigger role. And I think that even if you haven't explicitly formulated it yourself, any professional I think feels it, um, understands that that's the big, uh, play. So I think when it comes to people thinking about their skillset. And what they have to develop, even if they don't explicitly formulate it like that. I think it's mostly actually influenced by AI or the perceived threat of it. And, uh, as far as I've spoken to people individually, yes, I would say that the number one emotion that I get is fear. Like, uh, I, I would kind of put it in the fear category, fear of the complete unknown, and fear of like this technology being. Massively disruptive, right? Like more disruptive than any other that has come before. And there, there have been quite a lot of disruptive technologies in the music industry. I would put it that the music industry in this regards is one of the most violent industries I've ever seen. Like the disruptions that this industry has been going through basically every 10 years are just incredible. And that is what's driving people in terms of the view. They use skills. And it was kind of surprise to me. What I took, you know, as the most important takeaway is this multimodality out of the answers. Meaning most of the professionals are thinking that in the future they will not be doing their professional. You know, wellbeing is not dependent on just, you know, being professional in this one skill. They have to master variety of skills, and I do think that that's exactly. I mean, that would be exactly my perception of what AI does to the industry because AI is gonna enable people to become more efficient. And you will be able to do much more than what you are doing right now, which means that like the individuals that will be kind of top of the line in their profession, they'll be able to do a lot of things. And this multimodality means they, people will be better at communication, marketing, you know, technical skills, creative skills, so all variety of skills. So I think that. In this regards, the future seems to be the opposite of being highly specialized. So I think the, this is the kind of high level net takeaway for me is that at least that's what most of the professionals are currently thinking is that they have to go beyond, uh, just narrow specialty skills. I wonder if it would be interesting to learn more about people's attitudes towards generative AI in other fields, because it seems to me that if you're a musician, the prospect of generative ai, replacing your role is pretty scary. But the prospect of generative ai allowing you to make a music video without hiring. A whole team to do it is quite in appealing. And I imagine that the same thing works in reverse. If you're a video maker, the idea that you no longer have to hire a composer to make a soundtrack for your video is somewhat appealing. Yeah. Completely agree. Yeah. And, and one interesting thing that comes outta this survey as well is, uh, we asked. About the impact of AI in relation to different musical genres. And there seems to be a perception that the impact of AI is going to be quite strongly genre specific. People seem to think that genres like jazz and blues are quite resistant to ai, ification, whatever that means. Uh, whereas something like EDM or ambient music is, is much more vulnerable. Um, the thing I find interesting. There is that when I, I, I go on YouTube and I listen to AI music. What people are actually making with tools like Suno and so on, is I hear loads of blues. Right. And it's, it's quite convincing. So other results of the survey, uh, sort of outta step with reality here, do you think? Frankly, I do think that this aspect is rather biased, meaning I can just speculate of where the bias might, uh, come out from. I think that, you know, when you look at genres like jazz or blues, they, they are more, I mean, they are kind of the, the peak of the creativity, right? In a way of the human creativity, not in a, I don't mean it in any way kind of putting down other genres and saying that pop can be creative, but in terms of, let's say. You know, complexity of improvisation, right? Like, which many people will associate of this deeply humane skill of, right. Like a jazz band getting together and doing like a live improvisation, right? And that is, I think, perceived by many as this, you know, the complex aspect of creativity and, and it feels like the one that machines will be less likely to replicate. I frankly do not think that that is the case with machines. I think they will be actually pretty good at those complex genres as well. From a music theory perspective, complex genres, I think the machines will actually be as good. So the bias might come from the human perception that this complexity, that creative complexity is difficult to replicate. Uh, nobody knows for sure. I think that there is a bias there in terms of what we as humans, as a group, kind of perceive as complex or not complex for machines to replicate. So there might be a surprise in that, but your question. Uh, kind of drives me to another aspect that I would like to maybe kind of discuss a little bit in terms of, and that was what, what I've been thinking for the last few days is like, how does the industry look like overall in five years at the moment? And here's my take, like just, you know, from, from the last days of thinking at the moment, we are mostly in this new, like, streaming era, right? Like. We went from the CD era into the streaming era. We are, I think, fully kinda into that, uh, new paradigm, which is basically, you have all the music stored in the cloud, right? And then people can access all this music and listen to it. But the paradigm is still, we are creating albums and the individual songs and grouping them into playlists, and then people are consuming those. And the use case is still a lot of people. Choose genres or artists and then they via those selections, they get through playlists. I don't know, the playlist can be my early morning, but then it's gonna be comprised of certain, uh, songs. I think fast forward, like we will have. I see at least two distinct modes of how the music will be consumed. One is probably still gonna be the same paradigm in terms of songs created by artists and maybe grouped in playlists or albums. But the other is gonna be just purely real time generation based on the context, meaning if I wake up in the morning, then my musical AI agent agent knows, uh. Generally what I want in the mornings or what I want while I work and want to focus. And then it'll be generating the music kinda real time. And probably that actually gets away from the concept of songs of such, because why actually have the interruptions in between, right? Like if you are in the deep focus mode and you are working. Music in the background, uh, might just be a continuous flow without actual interactions in between. And that's a completely different mode of how to think about, uh, music consumption. And I think that sort of, we will have those, these two worlds will gonna emerge. So you will still have quite a lot, uh. You'll still have artists creating songs for the sort of, uh, the song, uh, sort of, uh, listening mode. And you'll have AI agents actually creating some stuff on the fly. So I don't know where the division stands in terms of which one is gonna be bigger and whether it's 50 50 or. Whether one is much larger than the other, but that I think very much changes how we might be even consuming the music. And this then comes to the question of, and that's an important question, if we want to further investigate the generative aspect of ai, who should we actually even ask the question? My feel is that the professional producers are maybe not the right audience for that. It's probably the consumer and trying to understand the. Mental patterns that the consumers have and how they would be, uh, sort of open or not open to that kind of listening. And then probably the distribution platforms, uh, meaning the current streaming platforms, maybe the big tech companies. And, and there you will instantly have some bias if you try to kind of get the objective answers, but I think those are the audiences we should be addressing in the future if we want to understand the generative aspect, uh, more in depth. Interesting stuff. Um, I was interested by what one thing you mentioned there about the possibility of generative ai, simply making the idea of a song redundant and in a sense that's just an extension of what happened with previous technological revolutions. Alright, like. You know, the reason we have three minute singles is because they fit on a 45 RPM disc. And the reason pop music is like it's today is because of TikTok and or streaming and so on. So it is not a radical idea in a sense that technology shapes music rather than the other way round. But that's pretty extreme example. Yeah. Yeah. This has been a continuous development, uh, meaning, you know, you had the album. Largely defined by the physical reality of the product, right? Where you were fitting them on the vinyl or on the CD disc, and there was a specific form of limitations for that. So you actually have an album concept out of it. But when you think in terms of, let's say, Euro back 300 years, I don't think that the Moer and V and whoever had the concept of albums, right, like the album and this way is up. Completely artificial creation out of the technological sort of limitation. And right now I think we are firmly in the playlist era. So I think the, actually the album era kind of somewhat went away with the C and there's still an inertia there to kind of think in terms of albums. But I think if I look at streaming, it's clearly an arrow playlist, right? Like it doesn't matter. What happens to your album? What really matters if you are in the right playlist? The right playlists are the ones that are actually propelling the listen counts. So yes. And then, uh, an interesting aspect. I have, um, quite a few of course, uh, you know, friends in the music industry and I have one that is organizing our locally, some of the biggest festivals that are happening locally and for some time already. His complaint is that it's very difficult. It's becoming difficult to market. Festivals because when he puts artists on the posters, most people don't actually recognize the artists. If you play the song, they instantly know the song. They know, oh yeah, this song I know. I just didn't know who the artist actually is. And this is part because people are listening to playlists mostly. And that's kind of, you put it on, you forget about it, and there is just music playing for you all the time. So people don't, and. People. That means like, you know, the masses, when you organize a large, let's say local festival, you don't have the A level stars in your festival. You put, you know, the, whoever you are getting in the festival. And people just don't know the artists, even though they know the songs, they know the music. So, and that's, I think is a clear example of, uh, changing that, that mindset in terms of the playlist actually being the most dominant thing Now. As an evolution of that, yes. I think it's the playlist also goes away, and then there is just contextual kind of music for the context that you have. We already have that reality in games, like games have contextual engines that generate music. They're actually, you know, pretty complex and I think they are really good examples of what's coming in the future. So I think that concept will apply to actual mass, uh, music consumption. Where ultimately the engine, you know, like in a game. I haven't played, uh, lots of games lately, but uh, back in the day I used to play and you know, you are walking somewhere in this game world and approaching some danger and then suddenly, right, like the music starts changing in the background and there is that your sense, the danger. So I think that's what's gonna happen. Uh. Uh, in the future, maybe dangers, learn to play like this, uh, alert, uh, music, uh, before like you have a car crash, right? Like you're driving in the car and suddenly the environment changes, so you know that you have to pay attention. Wow. Fascinating stuff. Thank you so much, Hels. It's been an absolutely engrossing conversation and it's been wonderful for Sound On Sound to be able to work with Sonar Works on this survey, which I think has generated some really interesting and important results. Uh, before we go, I, I want to leave you with one last question, if I may. We've talked a lot about how. The music industry is very prone to violent disruptions. These take place perhaps once a decade or something, some technological development happens and turns the industry on its head. And when this happens, it's tempting to assume that in some sense history repeats itself and we can learn from previous disruptions about what might happen with this disruption. Now my sense is that AI is not like that. The AI revolution is not like any of the disruptions that has happened before, and there's a limit to how much we can learn from the past. What's your take on that? I, I agree to that and, uh, I, you know, frankly, I've, of course I've been following this industry. There's one interesting news piece, which is I think extremely important for the industry, which actually took me by surprise. So there, there are the generative engines, right? Like suno and, and, and the rest of them. And naturally what happened was labels went after them legally, right? Like there are all these court cases that have been filed where labels are skewing. All these generative AI companies, and this is somewhat a repeat of a history, right? Like when the pirates came and threatened the industry, the c industry with the downloads, right? Uh, that was, I mean, the, the labels took an extremely defensive role and they ultimately were there. Sort of in this somewhat called denial phase for very long time until they almost, I think the industry was almost killed at the very, very last moment where I think there was no other way. They just turned their face towards the new kind of paradigm and started embracing, you know. Ultimately streaming. That's how we got through streaming. Uh, but the industry went, uh, down, it lost two thirds of right. It went from I think, 20 billion in terms of record revenues to something like eight or so, like a, a factor of three. And it took a lot of time. What surprised me, and I was expecting that the same pattern is gonna happen again, like labels are gonna gonna take an extremely defensive role. For a very long time, and just very recently, I think it was two weeks ago, we had the announcement that, uh, Warner signed the deal with Suno. Right. I was surprised by that, and I think that is, to me, that was not a repetition of the history, meaning that they seem to come to, uh, face the reality much faster than in the previous experience. Uh. I don't know, is it good or bad? And for who? Because it would be very important to, you know, understand from whose point of view. We're trying to look at the question, but at least this seemed to be like for, at least for me, stepping out of the historical pattern and the sort of incumbents, which I would say labels are the dominant force at the moment in the industry to come. Much more proactive role kinda in the dealing with this. So that is a small nuance in terms of not repeating the pattern and uh, yes, I think the essence of the AI is different. Uh, the biggest reason being that the previous disruptions were mostly formats. It were the formats that were sort of the most disruptive, like going from a CD to a download. And of course there was some change in the creative process in a way. As the formats changed, but those were kind of nuanced. You like these disruptions never affected the, the really core of the creative process, maybe with the exception of the onset of the recording industry itself, meaning at some point in the history, and I would be. I'm too young to kind of pinpoint the exact decade when it happened, but initially engineers were kind of completely left out of the creative process. And then at some point as these digital technology started to evolve, you had the actual recording, uh, engineers and mixing engineers. Taking a very active role in the creative process itself because technology became kind of interwind with the musicality of it. I mean, uh, the musician, uh, you know, just does the instrument playing and singing. But then there are all the technologies like reverbs and delays and effects that were put on. So engineers became like the part of the core process, and that was, I think an example where the technology shift was affecting the creative process. Then let's say a change from a CD to a download that's a pure format kind of play. AI is definitely, is less so a format and more so of the creative process change, but there's also a format. So I think in this respect. It's gonna be more disruptive in different ways than any of the previous generations. And I do think it's extremely difficult to predict where it actually takes us. But yes, I think it's gonna be, it's gonna be a different disruption process, and I don't think that history is actually a good guide here for what's to follow. We live in interesting times. Thank you so much Helmuts. You've been listening to the Sound On Sound recording and mixing podcast with me, Sam Inglis and with Helmuts Bems from Sonarworks. If you'd like to know more about the results of our survey, you can read an analysis in the February, 2026 edition of Sound on Sound Magazine. There's also an extensive white paper on the Sonarworks website. In the meantime, please do check out the show notes page for this episode and check out our other episodes and channels online. Thanks for listening.