The impact of artificial intelligence on society in general, and the music industry in particular, has accelerated. AI-powered tools are invading our private and professional space, and music is no exception — AI is present in many tools, from the recording studio to rights management.

But issues and questions surrounding AI have also multiplied, not least on the copyright front. Who owns works authored by AI? Should works used to train AI tools be licensed? Is AI going to take over work usually done by sound engineers, lyricists, arrangers, musicians, singers, etc?

Markus Schwarzer is CEO of Cyanite, a Berlin-based company that develops powerful and easy-to-use AI tools for music search, recommendation, and analysis. In the opinion piece that we are publishing below, he argues that the music industry needs to change its perspective on AI and see it as an object that can empower the sector, rather than undermine it.

The views he expresses are his own and do not necessarily represent those of Creative Industries News.

Artificial intelligence has become the bête noir of the music industry. They fear it lies in wait, threatening to make artists obsolete by replicating their signature sound, or generating engineered pop to hook the masses. The rise of the machines will crush creativity, leaving us trapped on an endless elevator ride of bland muzak.

But AI isn’t a sentient technology bent on taking over the world one catchy jingle at a time. This kneejerk reaction to its use in the music industry is rooted in a limited understanding of both its functionality and its limitations.

While AI-generating music is catching the limelight, in reality, used to its full potential, AI can be a force for empowerment and democratisation in music; levelling the playing field for artists, allowing publishers and labels to reimagine their back catalogues, and enabling streaming services to create next-level user experiences.

The theory of AI in music

Therefore, AI isn’t stealing musicians’ jobs; and content generation is just one of the roles it plays in the music industry. Behind the scenes, AI is working to improve access to music and understanding of how people respond to it. A deeper understanding of AI in music can help the industry appreciate the benefits it brings – and also recognise where its limitations lie.

Music creation is the most obvious application of AI in music, but the extent of its capabilities is often overestimated. AI can be trained to create custom music for specific contexts, to elicit broad emotions, and adapt it to changing situations. It’s a cheap and customisable way to produce new, rights-free content with Beatoven.ai, Mubert, and Boomy being the leaders in that field.

What it is not is a substitute for the complexity and emotional connection attained by human composers. Fully synthesized AI music is still low quality and limited in its influences; consequently, short, hybrid tracks, sampling pre-recorded and instrumental music for specific use cases, are perhaps the best product of its powers.

A useful role in the editing suite

Arguably, AI has a more useful role in the editing suite, providing an affordable, fast, easy to use mastering solution for those without access to high-tech studios. AI mastering offered by companies such as Masterchannel and LANDR allows bedroom producers and musicians creating demo tracks to achieve the sound quality needed to impress studio executives and meet streaming platform standards. It can turn a mediocre music file into a polished piece. However, it does generate a certain uniformity of sound and is lower quality than professional human mastering.

Where AI’s current most sophisticated and powerful talent lies is auto-tagging – a job that people will be happy to delegate to a robot replacement. Music libraries need to be categorised by useful search terms, so tracks have to be tagged not just with artist and title, but by genre, mood, tempo, language and energy, in order to be found. Human tagging is time consuming and subjective, leading to inconsistent or incomplete tagging and leaving content undiscoverable. But AI can be trained to identify genres and sentiments. It’s fast, consistent, objective, and gives the flexibility to retag almost instantly.

From better tagging comes more accurate search and recommendation, another area where AI can contribute massively to the music industry. With 100,000 new songs released online every day, standard search algorithms are struggling. But AIs use semantic search to pinpoint given elements within the complexities of music; essentially, identifying tracks which ‘mean’ a particular thing through sonic similarities. This provides a seamless listening experience. What AI can’t do is be human. Sometimes our mood is best expressed by mashing Mozart with Miley Cyrus – something an AI just wouldn’t understand.

Music AI in action

The real empowerment AI currently delivers to those in the music industry lies not in theoretical use cases, but in real-world examples. At Cyanite, for instance, we employ powerful AI recommendation, tag and search solutions to help music companies such as BMG, APM Music and Pond5 turn their catalogues into their own proprietary Spotify. This allows music, entertainment and advertising companies to deliver the right songs in response to customers’ search queries, however obscure.

For creators, AI offers a toolbox of practical music applications. These include audio transcription to create lyric sheets; source separation to pick samples out an original work; and mashups mixers that perfectly align beats. For the poetically minded, the use of natural language programming makes lyric and text analysis, and therefore song writing and ideation, easier, by drawing out common topics and sentiments. Voice synthesis can give songwriters who aren’t singers a way to get their songs heard by those that are.

Even though everyone is talking about AI generated music, the topic of discovering existing and human-recorded music is one of the most exciting and challenging elements of the music industry – finding the perfect piece for the perfect setting in the masses of today’s music. With its ability to search inhumanly large volumes of content, AI opens up the entire world of music – or at least the entirety of a given library – to music explorers. AI’s power for recommendation also guides fans to their new favourite thing. Rather than being restricted to a genre or radio station playlist, where only some of the stuff is for them, listeners can tell an AI recommendation application everything they like and it will deliver a playlist composed of all of those elements – and nothing else – which can then be cut by mood, activity, or who they’re trying to impress.

How AI delivers democratisation

As the music business is led by big players wielding a lot of influence, anything which can introduce more shared value is very welcome. In a more democratic model, more artists would have access to more opportunities, as would talented professionals on the technical side, working out of home studios or for tiny labels. By decentralising the power in the industry, more control goes to the creators – along with more money.

AI facilitates this democratisation by removing barriers and even revolutionising the traditional music industry model. It gives every single music creator the same chance to be listened to and loved; or picked up for a commercial, TV show or film, regardless of their popular profile. It hands listeners the power to choose exactly what to play, rather than being fed what the big labels, radio stations and streaming services want them to hear. It is even empowering musicians to have better legal backing and more easily defend their copyright.

At the risk of sounding idealistic, AI gives us the power to liberate music from the chains of commercialism, while elevating the value of creativity and quality. That’s something to be celebrated, not scared of.

By Markus Schwarzer, CEO of Cyanite

(Picture under license from AdobeStock)