Young composer and pianist Zhu Xuzhi (right) and audio algorithm engineer Gao Yuejie.

Photo by China News Service reporter Zhang Hengwei

　　China News Service, Shanghai, March 17 (Reporter Zhang Hengwei) On March 16, the first workshop of "Digital Muse-2024 Music Technology Sunac Festival" held by Cadillac Shanghai Concert Hall "Music Black Technology: AI War" Among human composers, who is the real god of music?" held at the Music Cube of Shanghai Concert Hall.

　　At the workshop site, young composer and pianist Zhu Xuzhi and audio algorithm engineer Gao Yuejie conducted a PK battle of "human composition vs. AI composition" using improvised "human composition" and "operating system AI composition" respectively.

　　Subsequently, pictures, text, musical notes and other content randomly provided by the audience are "fed" to the AI ​​for "artificial composition" creation.

At the same time, Zhu Xuzhi also used the "human brain" to improvise music clips with the same theme, allowing the audience to intuitively experience artificial intelligence redefining the boundaries of music creation and how it and human artists can jointly shape the future of music.

　　After two rounds of competitions, the AI’s performance amazed “human composer” Zhu Xuzhi, “No creation is accomplished overnight. After I finish composing the first four sections, I often get stuck on how the theme will develop in the future, but AI You can continue writing directly. Although there are slight traces of the machine, I can feel that it is understanding the music, which surprised me very much." Zhu Xuzhi said.

　　However, Zhu Xuzhi also pointed out: "AI composition is basically within a correct music framework, but to me, your aesthetics will be tired. When this kind of music becomes popular, everyone will miss art more, or be more aesthetically inclined." Something to pursue.”

　　Speaking of AI composition, how does it understand music? Gao Yuejie explained to the audience: "We humans understand music through two channels, one is listening, and the other is reading music scores. In fact, when AI understands music, it is almost These two methods. In the early days of computer development, only the data format of the music spectrum could be recorded. For example, 'listening to songs and identifying songs' used this technology. Now in the era of large models, through the 'feeding' of millions of data, Coupled with very powerful computing power, this kind of learning by computers is actually more like a process in which we humans develop a sense of music."

　　Although the learning path of AI and humans may be the same, Zhu Xuzhi believes that for a long time, it will be difficult for AI composition to surpass humans, “because what ultimately affects us is emotions.” Zhu Xuzhi said: “Although AI can It can learn all music theories and analyze music scores much better than us, but its selection is not based on emotional reactions. It is based on the models and basic theories it has learned, so AI can generate 60%, which I think is more correct. Music, but the last bit, the few notes that touch people, AI may not understand.”

　　Regarding Zhu Xuzhi's view, Gao Yuejie also affirmed: "AI is very smart but will not make mistakes. It is precisely because we have this ability to make mistakes that are constantly personalized that we have achieved many great works of mankind." (End)