The ongoing culture war regarding artificial intelligence (AI) and the arts is quickly becoming trite. There is much discussion about large language models (LLMs) and generative AI in the production of music. Adam Neely has an excellent video titled Suno, AI Music, and the Bad Future if you're curious, but I don't wish to join that conversation in this essay. Instead, I intend to focus on the other side of music: its consumption. How might AI and (hyper-)personalization algorithms affect our consumption of music?
The internet provided a powerful thing during the 90s: the ability to share files, music, and ideas to a much wider audience than ever before. Websites like AllMusic, MP3.com, and Napster were places where people could share and find new music; resources, such as Pitchfork grew to compete with popular magazines like Rolling Stone as the main medium for music criticism became cyber.
However, with the introduction and growth of music streaming platforms, there has been a decrease in the listener's agency over their music consumption. A critical step was algorithmized music selection; Pandora Radio was one of the first platforms that succeeded in this matter and hence is largely recognized as the first successful streaming service. Companies like Spotify and Apple soon followed and changed listening further with more personalization: personalized playlists (Daily Mixes, Daylist, etc.), personalized radios, and personalized algorithms for recommended tracks. While the nature of these algorithms remain the same, this listening experience uses far more of your data and is a large step up from Pandora's simple thumbs up/thumbs down.
Presently, the listening experience one can expect from streaming services, is one of hyper-personalization -- recommendations fed off of a vast amount of your listening data to a point where almost no novelty remains. Your entire listening experience is filtered through your past listening habits. On the surface this may not appear like such a bad idea; here are two criticisms.
(1) Loss of the new. Listen to music long enough on Spotify and you'll find that it consistently recommends you the same music. Novelty and originality are progressively more difficult to find in a system which is based on what you previously listened to. Most of us grow tired of what we listen to; we seek novelty, but even Spotify's genre playlists and artist and track radios were personalized for years, only recently adding a non-personalized option.
(2) Alienation. Through a hyper-personalized listening experience which is often repetitive, we lose our connection to others. Music should be a social experience, even in it's recorded form. If each of our listening experiences is nigh perfectly tuned to our precise taste, how do we discover music outside of that taste and how do we relate to others? A large part of music is the connections we make with one another; this is more apparent when making music together, but there is a kernel that remains in its listening.
The internet continues to evolve and becomes ever more entwined with the technology of LLMs, generative AI, a technology that fundamentally mechanically digests and regurgitates data ad infimum, a technology incapable of producing the new. These issues only exacerbate.
The growth of AI has and will continue to affect music in both its production and consumption. On the consumption end, LLMs are used to bolster personalization. This is unfavorable, but it is also moot; the streaming platforms already use ouroborosian algorithms. Algorithms will get better at predicting your behavior, but on its own this does not present a new problem. What is far more frightening occurs in music's production. Generated AI music (GAM) -- which has already infected most streaming platforms -- is being incorporated into this autophagic system. Now, not only is the platform recommending us 'music' which we are predicted to enjoy, but the 'music' itself is merely processed cultural past and offers nothing new.
The increase in alienation isn't much better. A surge of AI 'artists', which are then fed to you through AI-assisted personalization algorithms, a phenomenon we might call hyper-personalization, results in a complete loss of the new and of the real. It is a manifestation of Baudrillard's hyperreality and a postmodern breakdown of a shared reality -- certainly a place where it is difficult to discuss and criticize music. Participants of Suno cannot connect with one another since they only listen to their prompted creations and no one else's. Currently, Spotify is trying its best to deter GAM on its platform. I suspect it is not because they recognize the harm I have laid out above, but instead a more sinister motive.
Some speculation: Spotify, Apple, and other multibillion-dollar tech-corporations are already researching and designing Suno-like personalized GAM to integrate into their streaming platform. Indeed, while writing this piece, Apple announced Playlist Playground where users can use AI to create playlists -- a feature Spotify already boasts; Google announced that Gemini will soon be able to create 30 second songs; Spotify's R&D department lists "generative AI, LLMs, reinforcement learning, and causal learning" as particular interests. These are all just one step away from fully integrated GAM features and thus hyper-personalization. The end goal is to replace real artists on their platforms. Paying artists less mean more revenue right? Generative AI music and hyper-personalization will continue to grow while paid artists decline. Ultimately your algorithms will provide only slop, personalized to you of course, but lacking originality: insipid.
I am not opposed to using algorithms or even the reified 'AI' (with its respective rainbow graphic design) to find music, but the future I outlined feels all too plausible. I want the option to receive human recommendations, because for all the music I seek out, the personal recommendations from friends and acquaintances are always the best. Whether that is because my friends know me better than any algorithm or because having a close acquaintance's recommendation is an added bonus, I do not know and is irrelevant. I believe it is time to flex our ability to do that now, before it's too late. I believe we deserve the opportunity for the good old hey-check-this-out.