How AI Is Impacting the Music Industry In Positive And Negative Ways

The work done by engineers and computer scientists in Artificial Intelligence (AI) over the past few decades has been monumental. The ability to create music through programming has changed the music landscape. AI has positive impacts such as allowing musical artists to add different elements to their sound production and performances. However, the negatives are the limitations it can create in many job fields if it continues in this direction. Since this method is less cost-effective, it incentives companies of trying this method. Let's explore this further, and get a perspective on both sides of the spectrum.

First, as shown in Holly Herndon's 2022 Ted Talk, there are ways to enhance performers while also gaining exposure yourself. Holly allowed musical artists worldwide to add her voice to different songs and languages, using her software called HollyPlus.

Holly Herndon's program "Holly Plus"

This was not even seen as a possibility within a decade ago. Looking back into the history of musical software, the first computer to ever play music was a CSIRAC. It was seen as a success to play audible noises that resembled instruments. Now, songs that are generated through computers can be passed as music created by musicians and singers. In the next era, with computer AI advancing further, it’s important that artists are still involved in the process. Holly emphasizes this ideology in the Ted Talk, as she even had the musician Pher join her on stage to demonstrate her program. However, there are plenty of musicians and companies that could think differently. What if software similar to HollyPlus is used irresponsibly?

Examining the biases within the databases

Based on AI research, it’s been discovered that software can be based on biases within the databases. Dr. Edward Powley, an associate professor at Falmouth University, discussed this topic in a university lecture. He brought up how people of colour (POC) can be unfairly tracked through AI technology when the FBI and police departments locate criminals. Generative Adversarial Networks (GAN) specifically have been called out by media for similar seasons. In this network, there are two aspects that make it work similarly to a game. There is the generator, where websites can generate images of what's specified in the search engine. Second, there’s the discriminator which is an image classification network that identifies what images are real or fake. In some cases, it can be simple to see what is real or fake to the human eye, but it’s not always simple. AI can paint the picture without having all of the details. It can show a story without having any background information on it. It may not think for itself, with a conscious mind, but the bias in the data being transmitted gives off that impression.

Dr. Edward Powley speaking in a lecture

Relating these concerns to current music AI software, there are fears of marginalized communities being victimized. The AI is taking data from other cultures to assume how their music should sound. It may be manually imputed and programmed by a software engineer, but how can we be certain that the engineer is programming it as responsibly as Holly? These are the main concerns that can arise if there is no conversation about this topic. This is where the fear of AI originates when these questions cannot be answered directly. That’s why it’s important for music AI software to integrate machine learning into its process. As we know, machine learning helps discover patterns and allows programmers to make inferences or predictions based on these patterns.

Although it is a good option to integrate it into the system, it’s not going to be a saving grace. Machine learning is just number crunching and pattern recognition. The machine is not thinking, as it can't tell the difference between the patterns the programmer wanted to find and biases that might exist in the training data. It can’t follow a code of ethics or take responsibility for its own uses and actions. Finally, if the machine can’t take responsibility for its own uses, who will? Is it the programmer who designed it or the company that produced the software? That is the key discussion as we continue to head into this new era of AI technology. Although there are concerns and cautions, there is positivity surrounding Music AI software. The ability to play other people’s music and voices with their permission offer a whole new world that is waiting to be discovered.

References

1. TED. (2022). What if You Could Sing in Your Favorite Musician's Voice? | Holly Herndon | Ted. YouTube. Retrieved September 30, 2022, from https://youtu.be/5cbCYwgQkTE .