Why Are Music Companies Afraid Of Artificial Intelligence?

Music companies have been lobbying against the use of artificial intelligence (AI) to produce music. According to reports, music labels are urging streaming platforms to prevent AI platforms from using copyrighted music to train their AI models and create new music.

How Do AI Models Create Music?

AI models that generate music are trained with many hours of content, which is used to churn out new content.

OpenAI’s image generator and ChatGPT have already created waves with their accuracy in terms of churning out images and text, as requested by the user.

OpenAI also has a music generator known as JukeBox that can churn out a variety of music in different styles, and even imitate the voices of artists and create new songs.

While some generators depend on existing music and change the tune or voice, others can work on the user’s instructions about the type and style of music and create ‘new’ music.

The use of existing music to train an AI model is a legally grey area since the owners of the model rely on legal loopholes and semantics that exist within fair use policies to use the existing content.

Why Are Music Companies Against AI?

AI-generated songs have been uploaded on streaming platforms to the dismay of music labels who own the original work that was transformed by the AI.

The music labels argue that the AI only uses the artists’ existing intellectual property to generate content.

However, it appears that music labels might be okay with these modifications if the AI labs start paying up for using the content.

“We have become aware that certain AI systems might have been trained on copyrighted content without obtaining the required consents from, or paying compensation to, the rightsholders who own or produce the content,” said Universal Music Group, according to an FT report.

It is possible that in the future music, labels could strike deals with AI companies using their content for payments in return for the usage of copyrighted content.

While some of these AI platforms have shut down, others like Google’s music generator product have not been released due to fears of misuse by users.

Google also probably fears the fact that it could get sued by music labels or artists over the content produced by its product.

The product by Google has been trained on nearly 300,000 hours of existing music.

Behemoths like Google are usually careful about releasing controversial products as it could hurt their brand image, and cause legal trouble at a time when governments already are scrutinising Big Tech closely.

Why Is Copyright Such An Ambiguous Issue?

The use of copyrighted work for training AI models has been a concern for quite a while.

Currently, the “fair use policy” is used by AI researchers and companies to justify using copyrighted works to train models.

While it is generally agreed that fair use covers the use of copyrighted images to generate content, the end-use of the content generated by the model could play a role in deciding whether there is any copyright infringement.

For instance, making an AI model learn an artist’s style in order to generate pictures with that style could count as infringement.

The output of the AI could adversely affect the market for the artist’s actual art. On the other hand, using large sets of images that have no specific style, and using these to generate new art might not count as infringement.

Whether the work generated by an AI can be copyrighted has been a point of debate as well.

In the US, laws allow copyrighting of the work produced by an AI model if it can be proved that there was substantial human involvement in the creations.

So, if a user can prove that he has worked hard on the model to create a new piece of art, he might get a copyright.

Streaming helped nurse the music industry back to health after it was plagued by years of piracy.

Today, the music industry is reporting record revenues and profits. Yet, there are several issues with streaming that artist, streaming platforms and music companies are trying to solve.

These issues include royalty payments, copyright issues, AI-generated music, the use of ‘plays’ to calculate payments, and revenue distribution, among others.

Since the fair use policy does appear to side with AI researchers and companies, the AI-generated music issue might be a tough problem for music companies. 

This article was first published here.