CCB Magazin

David Stammer: "I see AI tools as a continuum of digital music production"

Foto: © Sebastian Weindel

David Stammer is an expert for AI tools in music production and also produces himself. In a conversation with us, he explains how and where AI is already being used in the music sector today and what the consequences are.
 

INTERVIEW  Boris Messing

CCB Magazine:Hi David, you are a lecturer at the Pop Academy in Mannheim with a focus on digital innovation. You also organize and host the Future Music Camp. Do you also make music yourself?

David Stammer:Yes, I also produce my own music. I started as a DJ more than 15 years ago, releasing mixes and beats on Soundcloud and making my first EP in 2021. I actually come from the jazzy hip-hop corner, but my music is developing strongly in a more electronic direction. I also have a small label in Mannheim with which we do events in the house area and produce records.

CCB Magazine: AI tools are increasingly being used in music production. Where are their areas of application?

David Stammer:There's a lot of that going on. Automated mastering with the help of AI tools already works quite well, for example. AI is also a component of many plug-ins in mixing, for example in equalizers. Then in the area of music generation, there are many plug-ins, but they are not quite sophisticated at the moment. I can implement those in Ableton or other music software: Google Magenta, for example, builds on a beat I've created, suggests drum pattern variations to me etc. The same can be done with melodies, where the AI suggests me alternative melodies or continues my melody. Loops of individual tracks or a loop of the master track can be analyzed by AIs and alternative loops suggested based on the analysis. AI is also used in speech synthesis and voicecloning. Of course, this is not necessarily legal. The most famous example is an AI-generated deep fake song that was passed off as a new Drake & The Weekend song and later deleted. But there are now artists who issue licenses to others to clone their voice and use it for their own tracks. And since 2023, the first text-to-music tools have come onto the market, where smaller snippets of music can be generated on a text basis. 

CCB Magazine:Can AI tools for mixing or mastering replace music producers or only make their work easier? How does this change the role of producers anyway?

David Stammer:I don't think the job of a mixing or mastering engineer will be replaced. The AI tools are a useful addition, but the sound of the mixing is still determined by the producer himself. What is changing, however, is the way music is made. AI can speed up work processes, but it can also inspire. Thinking a bit further, there could be even more serious changes: In text-to-music, for example, you can create parts of music through so-called text prompts as midi notes or audio. I tell the AI what characteristics I want for a particular melody and it creates it. If we think this development further, the work of a producer could change a lot, at least on a craft level.

CCB Magazine:For which areas can AI-generated music already be put to good use today? 

David Stammer:Anything that has to do with functional music works quite well. For example, background music in the lounge or music for advertising. But I think it's unlikely that AI will create entire pop songs; that wouldn't be particularly exciting either. Why would we want to listen to more boring generic music? Especially since listening to music is often associated with certain bands or musicians and their biographies, keyword storytelling. The distinction between art and popular music is very important here. And the possibility of entering into a process of co-creation with AI and thus developing new ideas and unprecedented art.

The more generic or schematic my job or music is, the higher the probability that parts of my job will be replaced by AI - i.e., anything that can be well automated or imitated

CCB Magazine:What about studio musicians? Does AI pose a threat to them? For example, could they replace a guitarist who has to record a part for a song?

David Stammer:If my only job is to be a studio musician, I could imagine that it will be more difficult in the future, especially with regard to the functional music mentioned above. On the other hand, if I have a certain standing as a musician, I will be asked by studios to play my instrument precisely because of my special style. Also, the vibe of different musicians playing together cannot simply be replaced. Technically, of course, it would still be possible. Then there's a saying I've heard from various industries: XY won't be replaced, but XY who don't use AI will be replaced. The guitarist, of course, can use AI to do his/her job faster or better.

CCB Magazine:More generally, what new jobs can be created by AI in the music sector? Which ones might be lost?

David Stammer:I would put it this way: The more generic or schematic my job or music is, the higher the probability that parts of my job will be replaced by AI - i.e. everything that can be well automated or imitated. On the other hand, I can well imagine that there will be new jobs at the interfaces between AI and music, in the professional handling of AI tools and prompts or in explaining and communicating AI in the music and culture sector - similar to the many new jobs in the music industry that have been created by social media, digital marketing and streaming. 

CCB Magazine:What influence does AI have on the marketing of music? How is it being used there?  

David Stammer:Music is currently mainly consumed via streaming platforms. Recommendation algorithms that suggest music naturally play a major role here, as they influence what is listened to. In this sense, AI is a very essential component of music marketing and music distribution. Moreover, there are streaming manipulations, for example by bots, that cause musicians to get less money from the overall pool of a streaming platform. This can lead to recommendation algorithms recommending AI-generated music, i.e. AI recommending AI music - and the 'real' musicians earning less as a result. Streaming platforms are trying to fight this, but it's not easy. AI is also used in various areas of music promotion, such as writing press releases, creating artwork or teaser visuals.

In principle, most developers of AI tools are careful to use only freely available data for training the AI, so that no copyright is violated

CCB Magazine:AI trains with large amounts of data. This raises legal problems. German copyright law only protects man-made works. What does that mean for AI-generated songs, which only create songs based on these human-made works with which they have been trained?

David Stammer:In principle, most developers of AI tools are careful to use only freely available data for training the AI, so that no copyright is violated. There are open source databases for this purpose that may be used freely. But in hindsight it is often difficult to determine whether copyright material has not been used - and that is a problem. The best solution would be for artists or labels to join forces with AI developers and develop the tools together with them. Or models need to be found where the artists whose data is used for AI training are compensated.

CCB Magazine:Are there already such collaborations between labels and AI developers?

David Stammer:There is, for example, an initiative by Universal and YouTube, who have launched an AI music accelerator together with various artists, including Max Richter. The interest in joint development is there.

CCB Magazine:To take a look at the big picture: With the introduction of the computer, the process of music production has been greatly facilitated. The same applies to the digitization of previously exclusively analog devices such as mixing consoles or effects units. In this context, how do you rate the development of music AI systems?

David Stammer:I clearly see AI tools as a continuity of digital music production itself. The digitization of music production has meant that you don't necessarily have to go to the studio to record music anymore. Also, you now have all the instruments in midi form on your computer or access to huge sample and loop databases. The AI tools are a logical extension and addition in this context.

Rubrik: Specials