The Ethical Concerns of AI in Music

You enter your car and open up your phone using Face ID. Next, with your voice, you request Google’s natural language processing system to show you the way to your destination. On the way, you listen to music on Spotify using their machine learning recommendation algorithm. Artificial intelligence is firmly built into our daily lives, and in a few years, you may be able to click a button and listen to a classical music masterpiece, or the next pop hit - except it will have never been heard or sung before. Advances in generative AI make the field extremely promising, and soon the music generator may be a device integrated in our homes like Alexa, Face ID, or Spotify. However, although the idea seems enticing on the surface, it does bring about quite a few ethical concerns. Firstly, AIs that produce music have to be trained on some sort of dataset produced by humans. This brings about the concern of whether or not using copyrighted songs to train an AI breaches copyright. Furthermore, if a machine could theoretically produce songs better than any human in any style, at any time, and in huge quantities, what would that mean for humans who practice music and art professionally and as a hobby? Although AI generated music should be able to legally use copyrighted songs to train without any catches, a concerning aspect of its advent is the fact that it removes a source of economic and spiritual gratification from an already consumer-based society.

A neural network, simply put, is a large mathematical function which requires an input and returns an output. However, in order to create the function, the network must be trained on vast amounts of existing data, usually produced by humans. In the case of generative music AI, neural networks are given large datasets of music created by a human, say Mozart, and then learn to reproduce similar-sounding musical pieces on its own. Since the massive collection of works by classical composers like Mozart and Beethoven are in the public domain, training a neural network on it brings few concerns. However, if one wanted to create more contemporary music, say pop or rock, by training a neural network on songs from Taylor Swift or The Beatles, the issue of copyright and trademark becomes an immediate concern.

Artists and their supporters have opined that since they spend an immeasurable amount of time, money, and effort perfecting their craft and producing the art that they present, and if a black box could theoretically take in their music and recreate something just like it and intended for the same audience, they deserve some portion of the credit for it, legally and monetarily. However, this line of thought fundamentally misunderstands both copyright law and the role of training data in a neural network. Training data is only seen by the network and its creators, and the music that it produces is simply the machine’s “best attempt” at creating output with a similar profile, which is then distributed to the audience. Similarly, anybody who listens to an artist and mimics or emulates their style is essentially a very complicated neural network taking in data through their ears, processing it with their brain, and recreating similar but unique output in a studio. In such a situation, almost no one would claim intellectual theft, and existing laws in the United States give no copyright protection to artists for their “style.” Therefore, the same treatment should be given to machines, which were in fact designed to emulate the way humans learn. Using an artist's songs to train a network is very much like using an artist's songs to train a human, and in that same vein, should not be prevented from having claims over the music it produces, as long as its output is sufficiently unique from the training data. Copyright laws are intended to prevent others from stealing intellectual property and distributing it without permission from the original creator, whereas in the case of AI, copyrighted material is limited to the training of the machine - it never reaches the audience’s ears. Instead, the output of the machine is a unique creation of the AI, and as such should be immune to claims of intellectual theft.

A potential solution to the copyright issue is simple. Without forfeiting rights over the music itself, creators of the AI can simply catalog or cite the artists that it used to train the machine, perhaps included in a description of the device. This way, the artists can retain their credit and popularity by being acknowledged, but still allow copyright rule to operate freely and as it is intended for the protection of the AI and its creators. Though artists may demand compensation or royalties for using them in a training set, copyright rule simply does not protect them in this situation.

Furthermore, apart from the issue of copyright, many fear the rise of AI-generated music to quickly outcompete and replace humans in art and spell disaster for arts in general, which used to be centered around human creativity and interaction. Since the beginning of industrialism, people have feared the adverse effects innovation and invention many have on humans and their role in society. The case is no different with artificial intelligence. For example, tech billionaire Elon Musk insists that in less than five years, “we’re headed toward a situation where AI is vastly smarter than humans... things will get unstable or weird.” The Tesla CEO, along with many others believe that humans have reason to fear a complete replacement of people on the part of artificial intelligence, and that may be no different in the case of music. At the moment, there are very few people making commercially viable music with solely AI, and even fewer who are actually making any money off of it. However, as with many other disciplines, it may only be time before technology catches up and starts to produce better-sounding quicker and in larger amounts than any single human or musical group can compare to. At the AI song contest, one of the competing teams asserted that soon enough, machines will be able to automatically produce marketable musical compositions which will potentially endanger a source of income for many people in the performing arts domain. After all, if someone can buy a box that generates beautiful new pieces at the touch of a button, they would have no reason to listen to a human artist who needs years of training, experience, and effort to produce the same results. Thus, humans may gradually lose their spot in the domain of the arts. This idea has been brought up repeatedly in the context of economic sacrifice - but people whose primary income comes in the form of art only constitute a tiny population, about 1.4% of the United States. Meanwhile, the amount of new jobs that AI will create is massive compared to that number, meaning AI-generated music is unlikely to cause disaster economically to artists in the long run.

However, one ethical concern that has hardly been discussed regarding the replacement of humans in art is one not of economic concerns, but one from a broader, perhaps more spiritual perspective. Even though the amount of people that perform art as a primary means of income is tiny, over half of the United States engages in some form of art, many times as a hobby or pastime. In that case, having humans no longer creating art and becoming reliant on machines for creativity and relief in life can be catastrophic for the human psyche. Hobbies, the pursuit of knowledge, and the desire to learn something are many of the things that we as a society hold onto in order to have a sense of purpose in life. Artists commonly feel like by producing art they are not only reaching their true potential but also attaining fulfillment and satisfaction as individuals with purpose and identity in life. In a WALL-E-type world where everyone lives solely for consuming and deriving satisfaction from the comforts of technology and innovation, humans lose their sense of purpose in life which in turn has a devastating effect on the human psyche and reduces us to beings only meant to survive and consume. Although I am firmly a believer in the pursuit of knowledge and innovation, I also believe that art is one of the few outlets through which humans can express their true inner creativity and identity, and a world where humans have no interest or bearing in the creation of art is most definitely a dystopia.

One way to prevent society from becoming such a world would be to insist on people keeping humans as a central element of culture and art. For example, universities, schools, and other institutions hosting live music and art showcases can be enough to remind people of the integral role and infinite potential humans have in the performing arts. In such a world where music generated from a computer solely to consume is the norm, we must realize that in order to retain what has been an essential source of satisfaction and enlightenment for thousands of years, we must keep humans in the center of the musical sphere.

Science is a double edged sword. The same technology that makes the threat of long-range nuclear missiles imminent is also what got mankind to the moon in 1969. Generative AI, though perhaps not as drastic, can have its own benefits and disadvantages. Although creators of AI technology should have the right to use copyrighted training data, we must make sure that humans do not, as a result, become replaced in the field of arts. As a society, we have reached a fork in the road and now it is ultimately our choice what direction we head towards - a society of creators or a society of consumers.