Artificial intelligence, creating music, and copyrights – how will AI change the music industry?

The situation is at the same time bewildering, exciting, and intriguing. I claim that none of us can entirely predict at this moment where this all will lead. But we are on the threshold of the next musical/technological revolution.

Since copyright is about money (among other, loftier, ideals), AI will face conflicting fears and expectations.

What’s AI got to do with it?

The web is starting to fill up with examples of songs made by AI created fake versions of well-known artists like Adele and Drake. Even David Guetta was excited after trying AI for the first time. As a result, a fake Eminem could be heard at the beginning of one of his recent shows. I’m sure there are differing opinions about the results. Having tested some of the services myself one thing is for sure. It’s very easy to create a simple output. At least compared to doing the same using more traditional methods.

The core issue is much bigger. Researchers at Google have released a demo, an accompanying research paper, and a dataset for MusicLM. The idea is like what we have already seen with images: in this case it’s music that the AI generates based on textual descriptions. It’s important to note that developing this type of AI requires existing data. MusicLM leverages music found on YouTube and includes copyrighted material.

It will still take some time for Google to release MusicLM to the general public. One reason is that a small part of the music generated by the AI has been copies of the works it was trained on. But it’s only a matter of time before this or similar technology becomes more widely used. As Alex Mitchell, founder/CEO of a music and AI-based startup, Boomy, stated in a Billboard article: “if you think 100,000 songs a day going into the market is a big number, you have no idea what’s coming next”.

What follows from that? At least the value of individual works on average will decrease as the number of available works explodes. There will still be hits that receive the bulk of attention and listening. It’s also possible that “made by humans” will gain a luxury status and have a certain added value. But at the other extreme, the easier it is to produce a work of music good enough for the circumstances, the more it will be used for music that fits a specific situation. Especially if it’s also seen as a cost-saver.

What should we then think about all of this? I have the perspective of a musician and music creator working at a copyright organization. And I have many questions. So far there seems to be more of them than answers. At least ones that are more than opinions.

How does AI relate to music copyrights?

Generative AI has the potential to blur the lines between content created by humans and machines alike. What counts as creativity in the era of AI? Is AI an author like humans? Does the use of AI affect the threshold of originality? Is it possible that AI will produce the same work for many users? Does the creator of an AI have rights to the resulting work? These questions must be answered to determine who has rights to the output of AI.

The need for new legislation

In the United States, there have already been legal cases about the authorship of AI. In a 2022 case presented by Smithsonian Magazine, the U.S. Copyright Office determined that an image produced by an AI developed by Stephen Thaler “lacks the human authorship necessary to support a copyright claim.” This type of legal case is not unique, and they will become more common.

Using AI for music creation raises other copyright-related legal questions. Who owns the rights if an AI creates something? Who is responsible if AI creates something that infringes on existing copyrights? And to what extent can the companies developing AI define rules in their own terms of service? To resolve these questions, new laws need to be enacted, at the EU level (from a Finnish perspective), defining the rights and obligations of both AI creators and users. The matter has been under preparation in the EU since 2018. But it will take some time before anything concrete comes out of it.

Protecting human creators

The rise of AI will disrupt the traditional creative industry.  As AI-created works become more widespread, it is important to ensure that the rights of human creators are adequately protected. For example, by making the above-mentioned changes to the relevant laws and regulations.

The role of copyright organizations

At Teosto and other copyright organizations, it is necessary to ensure that customer agreements, distribution rules, and other necessary documents are up to date. We must also have an active role in the discussions about the use of music and related data we represent when they are to be used as inputs for AI.

AI is here – it’s time to compose some rules

In this blog post, I have used the term artificial intelligence (AI) in its broad sense. But I do not claim that a program that produces outputs that seem creative and intelligent is intelligent or creative. At least in the sense that we understand human intelligence and creativity.

Also, I am not worried about machines replacing humans as artists. AI is a poor master, at least for now. But a great servant, nonetheless! This also seems to be the opinion of professional music creators interviewed for a Forbes article. For professionals, it will be just one tool among many. And to be completely transparent, AI did help me write this blog post. It did not replace research and thought work, but it sped up text production and English translation.

In this situation, Teosto and the music industry in general need to talk about and agree on the rules of the AI game. The cat, as they say, is already out of the bag!

---

Text: Antti Rask
Photo: AJ Savolainen