Interface News Reporter | Yin Qinglu
Interface News Editor | Yellow Moon
Following the controversy caused by generative artificial intelligence entering the field of painting, in May, the uproar of AI Sun Yanzi and other AI covers made the music industry the next big hit. In the B station video, Sun Yanzi sang Jay Chou's "Hair Like Snow", and Wang Xinling sang "Set of Horse Rods" strangely, some videos have reached millions of views, and thousands of barrage comments.
Sun Yanzi later posted a response on social media, believing that human beings cannot surpass AI is just around the corner, and for the future, it is "everything is possible, everything does not matter", full of the meaning of being unhindered and helpless. In the face of the rapid development of AI technology, everything is still vague and unknown, but Sun Yanzi's response opens a bigger thinking gap for us: what kind of future may artificial intelligence lead to in the music industry? Will it free up the creativity of artists, or concoct more cookie-cutter saliva?
Interface Culture (ID: booksandfun) synthesized existing analysis and interviewed two senior music industry practitioners to explore a preliminary answer.
Before AI learned to create, algorithms had disrupted the music industry
In fact, music has been leading the way in exploring new technologies. Nearly forty years ago, the French scholar Jacques Attali argued in his book Noise: The Political Economy of Music that music, similar in nature to finance or mathematics, is purely speculation of mind, and that music tends to apply the latest technology to expand the boundaries of creativity faster than physical art such as painting.
On-site tuner Deng Chenglong said in an interview with Interface Culture that the reason why the audience will be surprised by the AI Sun Yanzi is largely due to the psychological uncanny valley effect, from the principle point of view, using technology to simulate the human voice is similar to using a computer to collect the sound of a piano, and there is only technical progress.
Li Xingyu, who has produced albums for many singers and is also an independent musician and sound space designer, mentioned in an interview that the application of AI in the music industry has long been very extensive, and many Internet celebrity songs are analyzed with relatively basic big data and produced by combining human resources. According to Li Xingyu's speculation, due to the rapid pace of technology iteration, a group of "quite decent" AI creators will appear this year. In January, Google released MusicLM, known as the "music version of ChatGPT", which can freely mix and match different types of styles and instruments, and by entering words such as "jazz at dinner" at a specified place and time, the machine can create music that matches the mood of the moment.
However, relying only on the computing power of AI, there have been no real hit songs so far, which is also a key node in Li Xingyu's eyes: "If AI can make a hit, companies and the most mainstream musicians who make a living from making Internet celebrity songs will definitely be hit hard, and the music industry will face restructuring." "In contrast, niche independent musicians have been less affected. This begs the question – if functional boilerplate music and saliva songs can be handed over to AI, can humans free up productivity and create more high-level original songs?
The idea is beautiful, but it is difficult to determine whether the reality will work out. After all, algorithmic AI has disrupted the music industry before generative AI learned to create. Chokepoint Capitalism: How Big Tech and Big Content Captured Creative Labor Markets and How We'll Win Them Back argues that in the early days of digital technology and the democratization of the Internet, power once rarely flowed from record companies to artists, but it did not last long, and now streaming platforms replace record labels. Grabbed most of the value of the music.
According to the book, a significant phenomenon brought about by streaming media is the "playlist culture", where listeners become children in need of care, preferring to passively listen to algorithm-recommended music rather than explore independently. The listening scene is also constantly subdivided, and there are even such precise lists as "hip-hop songs for cycling" and "breakfast in bed". Because playlists have become so important, excluded artists are likely to face song release failure, so they must keep creating tracks that are pleasing to the ear and have the potential to go viral, which tends to result in more mediocre, cookie-cutter compositions. On the other hand, in order to allow listeners to listen to as much music as possible and thus increase platform revenue, mood music playlists called "Chill Vibes" and "Peaceful Piano" have emerged in large numbers, they are not aesthetically challenging, and they will not get tired after listening to it all day, which is the kind of music that artificial intelligence is good at.
To add insult to injury, as streaming platforms allocate royalties from a limited pool of cash, the more a song is played, the more creators earn, and some creators resort to cheating, that is, entrusting third parties to let robobots listen to songs, further squeezing the living space of artists who do not cheat. A study by the French National Music Center found that 1%-3% of music on various platforms in France in 2021 was listened to by robots, which translates into a total of 1 billion to 3 billion times.
Li Xingyu said that the situation at home is worse than overseas. In recent years, the domestic music ecology has not been good, streaming media platforms push most of the resources to traffic artists, and the remaining resources rely on independent musicians to fight and fight, which has led to only the head and bottom in the industry, and there is no longer a middle layer. Coupled with the explosion of music variety shows during the epidemic, independent musicians need to try to reach the top with shows such as "Summer of the Band", so the performance market is also not healthy now, and audiences are either flocking to large-scale concerts like Mayday or going to see bands that have become popular in music variety shows - New Pants or Five is the best example, they are also excellent, but without exception have strong entertainment attributes.
In this context, even if the technology of AI composition improves, it is not necessarily good news for the market. First of all, for listeners whose listening tastes have been shaped by algorithms, do they need music beyond cognition? On the other hand, the allocation of resources is also a problem, Li Xingyu pointed out pessimistically in the interview: "The most likely situation is that the platform will distribute resources to its own AI, while the head artist will still ask for his share, and in the end there will be nothing left for other creators." ”
Can artists take back AI? The issue of copyright and the identity of sound
Li Xingyu's concern may be correct. The big record labels certainly won't miss this emerging market, Universal Music Group signed a black virtual rapper FN meka in 2019, in addition to the voice from humans, the song's melody, rhythm and lyrics are all generated by AI, and in its heyday, FN meka can get more than 500,000 Spotify subscribers and more than 1 billion TikTok views per month.
Ironically, Universal Music Group is also anxious about the invasion of artificial intelligence because it holds the song rights of superstars such as Bob Dylan, and its executive vice president Michael Nash has said, "AI-generated music is diluting the market, infringing on the legal right of artists to be paid for their work." ”
Who is really hurting artists' rights? A previous artist interview with Interface Culture pointed out that copyright is increasingly becoming a system to protect vested interests rather than niche creators. The article "Creating Odd Jobs, Widespread Poverty: Starting from the "Stop and Change of Tide" also said that the logic of the technology industry is no longer to invest in artists, but to attract users by providing content, and squeeze creators with extremely low-priced streaming fees; Instead of ushering in a bright future where everyone is an artist, the lowering of the technological barrier puts creative workers in a more competitive and insecure situation.
It is at this moment that the two-sided nature of technology comes to the fore. Dissatisfied with the monopoly of big platforms, some artists have begun to actively use AI to explore new economic models. One of them, electronic music artist Holly Herndon, argues that there are many dystopian narratives about technology right now, but technology can also be an opportunity if artists can take ownership of training data. Holly's approach was to deeply fake her voice, use it to sing songs in different languages, and license it to a closed network of collaborators in exchange for a revenue share of subsequent works. Her point is clear: It is artists — not companies — who should decide the use of AI in music.
Interestingly, rather than "using" AI as a tool, the idea of artists independently cloning voices is closer to a game revolving around identity. Deng Chenglong mentioned in the interview that when thinking about the attribution and copyright of sound, we can also change the angle, AI Sun Yanzi simulated not Sun Yanzi's real voice, but a completely two-dimensional existence, "but it also has its own reality." This leads to confusion and confusion, if the simulated sound does not belong to the artist himself, who does it belong to? And, to what extent does sound belong to the individual?
This sense of confusion and confusion is one of Holly's purposes for deepfakes. She cites musicologist Nina Sun Eidsheim as arguing that sound is public in nature. For example, the human voice is not a unique entity, but part of a continuous material realm; The singer's voice is made up of a series of body organs, the acoustic conditions under which the sound is felt, and the styles and techniques that the singer usually involves when training, in short, this is closely connected to other things. Holly believes this is important to rethink copyright, as she said in an interview:
We live in a punitive IP environment, which is not suitable for the development of the music sector. Because weaker people are constantly copied, and people who could have contributed to the industry can't make a living, the most persuasive new ideas are often marginalized.
Another pop musician who has adopted a similar strategy is Grimes. In April, she announced that she would allow people to make and commercialize music using an AI version of her voice, sharing 50 percent of the mastering royalties with each other. In fact, Grimes often reflects on the relationship between humans and machines in his works, and also imagines the terrible end after AI will rule the world one day, the song "We Appreciate Power" envisions a girl group dedicated to serving AI dictators, they sing and dance to offer flattery and kindness, and the lyrics also say "swear allegiance to the world's most powerful computer, simulation technology is the future"; One remix album is more direct, with a simple and crude title: This story is dedicated to all those cyberpunks who fight against injustice and corruption every day of their lives!
Grimes' imagination is not ridiculous. As mentioned earlier, technology often falls under the profitability goals of big companies, which makes technology itself seem "evil." But the evil is not the technology itself, Deng said at the end of the interview that he still hopes to remain optimistic, because according to past experience, every technological progress in the music industry is a good thing. He believes that AI in music is neutral, it can directly transform aesthetics into productivity, or it can become a tool for plagiarism and laziness, and it remains to be seen how it will develop.
Resources:
Noise: The Political Economy of Music by Jacques Adali, translated by Song Sufeng / Weng Guitang, Henan University Press, 2017-9
Chokepoint Capitalism: How Big Tech and Big Content Captured Creative Labor Markets and How We'll Win Them Back. Rebecca Giblin, Cory Doctorow, Beacon Press, 2022-9
Creation of gig work, widespread poverty: starting with "stop at station B"
https://www.jiemian.com/article/9224254.html
The plagiarism incident of teachers at the Chinese Academy of Art has exposed the problems of the domestic art system Interview
https://www.jiemian.com/article/9333500.html
This Singer Deepfaked Her Own Voice—and Thinks You Should Too | WIRED
https://www.wired.com/story/holly-herndon-ai-deepfakes-music/