Music has long been utilized as a means of emotional expression for humans. Artists take feelings of sadness, anger, and joy and transform them into lyrics and melodies, sharing their compositions with hopes of touching others experiencing similar emotions. Thus is the beauty of music — it allows for the emotion present in musical phrases to convey ideas to other like-minded people, creating a wordless exchange of thoughts.
With the growing development of new technologies like AI, jobs once filled by humans are replaced with highly optimized, well-trained algorithms that run at remarkable efficiency. AI’s impact has been monumental, with industries ranging from healthcare to transportation being vastly improved with the continual development of more effective algorithms. And it’s nowhere near its peak level of performance — AI is growing, and growing fast.
Researchers have been developing numerous algorithms that explore a multitude of fields, one such field being music. Projects such as MuseNet by OpenAI use powerful deep neural networks to create four minute, computer-generated music compositions. Trained on an enormous number of MIDI files, MuseNet has the ability to utilize 10 different instruments in its compositions, and is able to replicate and incorporate the aspects of various genres such as country or rock. While it is extremely impressive, does it really hit the same as human-made music?
As aforementioned, much of the joy associated with listening to music is the ability to understand the artist’s emotions. As advanced as AI are, they aren’t usually intended to be emotionally-intelligent programs, but merely computer algorithms that are highly optimized to perform tasks at high efficiency. And while there are AI that are able to comprehend and respond to the emotional cues of humans, AI lack the ability to develop and express their own emotions. It’s not very easy to transform the emotions of humans into a manner capturable by an algorithm, creating a significant barrier that hinders progress in AI with high emotional quotients. Thus, AI like MuseNet aren’t yet at the level of complexity to capture the depth of music in the same manner that humans can.
However, it can be argued that this is a positive aspect of AI-generated music that makes it unique. The ability of AI to generate music in a manner that is separate from conventional lines of human thought allows for current creative boundaries to be pushed, Vice writer Kaleigh Roberts states. This could lead to new advances in fields like music theory or even incite the development of new genres of AI-generated music. Thus, lack of human influence can lead to the exploration of new musical frontiers, leading to the discovery of musical concepts that humans could utilize in their own compositions.
With the current rapid growth of AI, it doesn’t seem to be a far-fetched idea to consider the possibility that AI will one day be able to develop an emotional quotient on par with that of humans, giving it the ability to infuse its musical creations with the emotion that makes human music so enjoyable. But, as of now, emotional expression exists as a chasm that divides AI-generated compositions and human-made ones, with new technologies being developed daily that slowly bridge over this large gap.
Sources:
https://mitsloan.mit.edu/ideas-made-to-matter/emotion-ai-explained
https://openai.com/blog/musenet/
https://www.vice.com/en_us/article/qvq54v/why-is-ai-generated-music-still-so-bad
