Artificial intelligence will change the music industry

-"Just as with the advent of the internet, artificial intelligence will change the entire value chain. It creates new issues, not least for music creators and rights holders about how they are affected by new tools for AI and AI-generated music," says Ankit Desai.

‍

Scanning theweb for talent_200D↩Scanning the web for talent_200D↩

Desai is the founder and CEO of SNAFU Records. Based in Stockholm and Los Angeles, the label has made a name for itself by scanning digital platforms such as Spotify, YouTube and SoundCloud to discover new musical talent. Desai himself has no ambitions to use AI to create new music, but others do and they use tools like Voicify and Uberduck to do so.‍.

"JohnnyCash sings Barbie girl"

This has resulted in curious soundtracks of Johnny Cash singing "Barbie Girl", Frank Sinatra's "Gangsters Paradise " - or "Wonderwall" featuring the somewhat unlikely duet pairing of Donald Trump and Joe Biden. These may be mostly comical, but they also demonstrate some of the possibilities of AI.

‍‍‍

Trainingon music that already exists

By using huge amounts of existing music in different styles and genres, the machine learning algorithms can be trained to recreate chords, melodies, rhythms, instruments and voices from different artists, creating entirely new music in the same style. Desai believes this will make it easier for everyone to create music. ‍

Make music production accessible

-"If you look at the Billboard charts, for example, the average song has 3,4,5 songwriters, because to make songs that are that good requires you to be good at a lot of things. You have to have a good producer, be good at lyrics, melody, singing, and it's hard to be a world champion at everything. But with AI, you can replace two or three of those roles, so it becomes much more accessible for people to create great music. I think that's where the biggest change is going to happen," says Desai.

Music consumption becomes personalized

The technology also makes it possible, as the examples above show, to tailor music to the individual consumer's wishes, whims and mood. This is not just a vision of the future. It's possible today," Desai explains.

-"The only thing missing is legislation. It's all about the law. But otherwise, all the pieces of the puzzle are in place. Where we're heading now is that all listeners, all consumers, can have a completely personalized experience. Say you like Karpe and you're in a bit of a bad mood, AI can generate a brand new up-tempo Karpe song for you. Or 100! But then we also need to make sure my buddy, Chirag in Karpe gets paid for it...

An uncertain legal landscape

At present, it is uncertain what specific rights Chirag has if someone uses AI to create new Carp songs. "New technology offers many opportunities, and with them come new legal issues," explains Silje Strandengen, associate at CMS Kluge. She specializes in copyright law.

-"One of the issues I've looked at is the use of data for training AI models, i.e. using copyrighted material in these models. There is also a lot of ambiguity around the use of AI in a creative process more generally. Does anyone even have rights to what is created? And in all of this, there are many sub-issues and a complex legal landscape," says Strandengen.

Protected data basis

In other words, the challenge is not just the music created with AI, but the underlying data.

-"You need large amounts of data and you need relevant data of good quality to train these AI models. The problem arises if this data is copyrighted in the first place. Artists are then afraid that their works and material, which they have created or sung, will be used in such a training process. This kind of technological creation process means that others can use AI to generate a large number of songs based on your material," says Strandengen.

Mass plagiarism or inspiration?

The problem is not unique to the music industry. Recently, Norwegian authors were outraged when it was revealed that several of them, including Jo Nesbø, Karl Ove Knausgård, Maja Lunde, Åsne Seierstad and Nobel Prize winner Jon Fosse, have been used to train AI language models.

One objection could be that machines do what humans have always done, both when it comes to literature and music: Take inspiration from the existing and produce their own.

-"Many people compare AI and neural networks with how we humans are otherwise inspired by the art and culture around us. You could probably argue that this network is inspired in the same way as we humans are, but at the same time I think there is something about the technology here that causes differences. For example, it is assumed that digital copies of the data are stored when the input is fed into the technical system. Humans will never have the capacity that a machine has, so. The way it learns and the amount of music it can create makes this a little different," says Strandengen.

Waiting for the EU

She understands musicians who are concerned that AI could take over their creative processes. Or if they could risk new music being created based on their work without being compensated for it. At the same time, she emphasizes that AI, as it is used today, is always human-driven and that the AI model itself does not take the initiative to create either music or art.

-"Many people are calling for guidelines and rules. We currently have no direct regulation of AI, and in Norway we often wait to see what happens in the EU. For example, there may be court cases from the European Court of Justice, which will say something about the assessments to be made. Or regulations may be introduced that Norway also implements. A lot is happening both within the EU and nationally when it comes to artificial intelligence in general. The fact that many people feel that the regulations are lagging behind is basically nothing new when dealing with rapidly developing technology.

Gramo is on the case

The use of artificial intelligence also greatly affects Gramo'swork to safeguard the rights of Norwegian artists, performers and producers.

- We are actively working on legal issues related to AI and copyright, and are involved in European forums, both on the producer and performer side, to discuss how this should be handled in our industry. We look forward to a common European regulatory framework being put in place," says Kim Skarning Andersen, CEO of Gramo.

‍

Automated work processes

The use of artificial intelligence is also on the agenda internally at Gramo, to look at opportunities to automate demanding work processes, for example when reading and interpreting large amounts of data where our job is to link recordings, rights holders and royalties together.

- We spend an enormous amount of time and resources on finding the right licensee, and currently pay out more than 95% of the remuneration to individual licensees. With more than 50,000 rights changes in our systems every year, it's interesting to see how artificial intelligence can be used to automate some of this work. That way, we can free up capacity for tasks where we need people with heads and hearts," says Andersen.

A time of change

-"We are in a time of transition, with great uncertainty, and developments in artificial intelligence are happening extremely quickly. We must therefore be aware of this, build expertise and work actively with the industry and legislators in the time ahead," he concludes.