Text | Wang Ruoye
When everyone is struggling with whether the Chinese music scene is surnamed "Hua" or "Zhou", AI has quietly given the music scene its own surname.
Previously, Chen Shanni used AI to complete the new song "Teach Me How to Be Your Lover", which was so fake that iron fans were surprised that "how can even the ventilation sound when singing is so realistic"; Later, Sun Yanzi's fans were self-sufficient, using AI to simulate Sun Yanzi's voice to cover "Hair Like Snow", "Mercury" and "Always Been Quiet", and Station B UP commented that "the hottest Chinese singer in 2023 has changed from Sun Yanzi to AI Sun Yanzi".
With the innovation and leap forward of artificial intelligence technology, all walks of life cannot escape the baptism of AI. Also in the field of art content creation, AI painting has been landed and applied to the actual industry, resulting in a large number of original artists losing their jobs. When musicians see this scene, it is inevitable that things will hurt them.
Chen Chenchen, a domestic artist and music producer, told Bone Duo, "Like painters, it is only a matter of time before a large number of musicians lose their jobs. It's just that at present, technology companies have not yet set their sights on the music scene. "It can only be said that music does not make money and saves the lives of musicians.
antagonism
The confrontation between musicians and AI is not overnight.
In 1956, the term "artificial intelligence" first appeared on the stage of history, and the following year the first computer-generated song, "The Iliak Suite", was born. After that, AI experienced a quiet period of more than 20 years in the music field.
Until 2015, the AI music startup Juckdeck was established, and various related companies and music works entered a concentrated explosion stage. In 2016, AIVA, one of the most influential AI music companies to date, was founded, and its music works began to be used by film directors, advertising agencies, and game studios.
The starting point of domestic AI music can be traced back to 2017, when the first domestic AI music startup DeepMusic was established, filling a long-standing gap in this field in China. In 2019, the Central Conservatory of Music also began to recruit doctors of music artificial intelligence, and the door to domestic AI music was completely opened.
In 2023, AI music is once again facing a "turnaround" due to ChatGPT's open ground thunder. Represented by Chen Shanni, the musicians who took the lead in using AI technology to complete music works continued to speak out on the stage, discussing the advantages and disadvantages of AI to the music industry.
On March 14, 2023, Chen Shanni released a new song "Teach Me How to Be Your Lover". Her fans almost praised the new song after listening to it, and the singing standard of this song was the same as before, even slightly higher than Chen Shanni's previous state. At this time, everyone did not know that every breath and every singing in the song was interpreted by AI.
Although Chen Shanni later said in public that during the production of the song, she did enough work to tune AI singing, and these workloads were not lower or even much higher than singing in person, the whole network was still in an uproar. Musicians have taken the initiative to accept AI and let AI "replace" themselves.
Chen Chen Chen told Bone Duo, "It cannot be said that AI Chen Shanni will replace the real Chen Shanni. Because before "Teach Me How to Be Your Lover", it was the real Chen Shanni who created her own musical style and laid a fan base through countless creations and singing. If there is no real Chen Shanni, then AI Chen Shan Ni will also lose its meaning.
Including AI Sun Yanzi, AI Jay Chou, AI Eason Chen, these popular AI covers on station B just satisfy the "selfish desires" of fans, if Sun Yanzi, Jay Chou, Eason Chen are not well-known musicians, these covers themselves will not have any symbols. "I've never heard of anyone trying to cover a song with AI Tenger because Tenger himself has been covering it. Using AI technology to achieve singer covers is worthless to some extent. ”
At the end of 2022, when Jay Chou was asked if he was worried about being defeated by AI, he also said, "Although AI can do a lot of things, it can't replace my beauty in music creation." ”
When the use of AI technology in the music scene still stays in a large area of tuning, cover songs, etc., the confrontation between musicians and AI is, in the final analysis, the public's identification of the emotional nature of music. What musicians can do is to convey the sadness, anger, joy and joy of creation through music. AI is not destined to have emotions, but musicians and music can.
coexistence
Although AI technology is becoming more and more mature, the concept of AI music is still not clear enough in the eyes of the public, which has led many people to mistakenly believe that the purpose of AI music is to completely replace musicians. That's true in theory, but it's not so when you look at it in detail.
The OXYZ3 team led by Chen Chen Chen is committed to creating a "virtual music world". Through this AI technology, listeners can "enter" the minds of musicians and actually see what they think when they are creating.
(Image source: OXYZ3)
"When we think about a song, we have a sense of picture in our mind, but this sense of picture is not present at a music festival or a live house. A lot of things can't be achieved in the physical world, no matter how much money it costs. "But with AI technology, audiences can see the performance in the musician's mind." Think of this virtual world as a documentary, MV, or other readable image. ”
After ChatGPT became popular around the world, the OXYZ3 team also reduced some personnel, which may also be one of the symptoms of "musician unemployment". Chen Chen Chen said that AI is the most helpful for small companies like them. "The work of collecting data and materials can be completely left to AI, and our efficiency has been greatly improved." As a user of technology, the company requires every employee to be proficient in AI-related knowledge. ”
The OXYZ3 team originally required one front-end worker and two back-end workers. But nowadays. Front-end workers can write back-end code directly using ChatGPT, so the team structure is adjusted to have one front-end and one back-end staff. This means that an employee of the music technology company lost his job due to AI.
But AI does not have the ability to work independently, and it always needs a person to lead him. Chen Chen Chen said, "The premise of writing code with ChatGPT is that you have to have a technical director who can read and understand the code. Otherwise, the code you write is unusable and needs to be modified.
On the other hand, the essence of AI music is to analyze and learn music data through various algorithms to form a relatively aesthetically determined style model, and then generate content in the selected style model based on user input. These models also rely on human use, not AI to use AI.
With the first place in the artificial intelligence music major, Zhou Mubei, a graduate student enrolled in Sichuan Conservatory of Music, introduced to Bone Duo that the song is divided into four parts to AI music: songwriting and singing. AI lyrics mainly use GPT-2 language models based on pre-training and end-to-end generative model frameworks. AI composition mainly uses generative models based on music theory rules and machine learning; AI composition mainly uses generative models based on music theory rules; AI singing mainly uses end-to-end acoustic model + neural vocoder.
Her current research is mainly studying MIR music information retrieval, such as note start point detection, chord tonality detection, beat detection, detection of underlying audio features, and then database construction. "AI may make us more passive in accepting information in the medium of music."
In other words, AI models are used to collect what kind of songs the public likes to listen to, and then use the collected information to create a model music library, and the music library will continue to produce new songs that meet the public's taste. "You can understand the music produced by AI at this stage as 'saliva songs'." Chen Chen Chen said.
Before AI Music became a "musician", it could only be regarded as a search engine, knowing technology, following likes and dislikes, and being able to sing, but not having the ability to work independently. Musicians and AI music have not yet reached the point where fire and water are incompatible, or it can be said that they are coexisting in friendship.
Overcome
Inevitably, Bone Duo asked Chen Chen Chen and Zhou Mubei respectively if they were worried that they would lose their jobs in the future due to AI music.
One is the founder of the first technology company in China to contact AI music, and the other is a vigorous student in the AI music industry, but their answers all point to one direction, "I don't worry about losing my job because of AI music, but look forward to the rise of AI music." ”
Although Chen Chen Chen believes that the unemployment of musicians is only a matter of time, he thinks more about how far AI technology can help the music industry develop. "The problem that has emerged in the Chinese music scene today is not the problem of insufficient creative volume, but the problem of insufficient production capacity of high-quality works. Yes, even if 1,000 songs and 2,000 songs are released, everyone is still listening to old songs and covers. Even music variety shows are mostly covering golden songs. Chen Chen Chen said that the overall vitality of the music industry is declining.
At present, AI technology is also unable to solve this dilemma. Whether it's ChatGPT or other human-only technology, these tools can only allow musicians to create faster. "For example, I could only release one album a year, but because of AI to help me write and sing, I can now release 10 albums a year. But can releasing 10 mediocre albums bring any substantial changes to the music scene? ”
In Chen Chenchen's view, AI music is actually providing great help to the game industry. The game industry needs massive music to fill game content, and music, as a carrier related to game screens, is in great demand but not very demanding. The production speed and midstream quality of AI music can perfectly adapt to the needs of the game industry.
"For musicians who really enjoy music, they are less satisfied with the work produced by AI. Musicians must make songs in the hope that through songs, they will have an emotional projection of the world, hoping that someone can understand the thoughts spent in the songs. Chen Chen Chen explained, "But after using AI to make a song, you turn to someone else after listening to the song and tell him that I wrote this thing with AI, no matter whether the person listening to it says that the song is good or bad, you can't enjoy the emotional value of this good and bad." ”
And for music fans, listening to songs is like going to a restaurant to eat. We must hope that the dishes served are carefully prepared by the master on the spot, not microscopic from the microwave. Compared with the creation of musicians, AI music is like convenience bees VS Michelin.
Zhou Mubei is grateful that AI has brought her to the music industry. She is a science student who has no conditions to learn musical instruments since she was a child, and it is precisely because of the establishment of the artificial intelligence music major that she can begin to carry out in-depth learning in the direction of audio music. It can be said that AI music lowers the threshold for individual creation.
The maturity of AI technology has made "taste songs" a reality. Even if you do not have advanced knowledge of music theory, are not proficient in any musical instrument, and do not know how to write and sing, you can make a highly completed song according to your own musical taste through an AI model.
In any case, the wheel of history rolls in, and AI technology will not stop progressing because of the fear and "opposition" of a certain group, you know, when Faraday created the generator in 1831, people also called electricity a monster. Musicians, as well as all of us, may as well boldly embrace AI music, accept, coexist and overcome.
(Note: "Zhou Mubei" in the text is a pseudonym)