After the explosion of "AI Sun Yanzi", where does the music go?
"AI Sun Yanzi" came out of the circle strongly and became the new top stream. Behind this is the unstoppable trend of AI music that has developed for more than half a century. After the intervention of AI, it has brought a huge impact to the music industry, and the biggest problem is copyright disputes. In the face of shocks, the government is supporting new technologies while constantly improving supervision, enterprises are advocating copyright protection, upgrading service experience, and promoting industry norms based on their own interests, singers are trying to use AI to assist creation, and the public prefers to play memes. In addition to worries, we may be more likely to actively embrace change, and AI intervention is expected to usher in a new era of music.
Author | Yan Xi (Author of Cultural Industry Review / Research Fellow, Sanchuanhui Cultural Tourism Research Institute)
Editorial | time
Edit | peninsula
Source | Cultural Industry Review
Recently, the "unpopular singer" "Sun Yanzi" returned. Note that this new top "out of the siege" is not Sun Yanzi herself, but "AI Sun Yanzi".
"AI Sun Yanzi" has amazing productivity, and the covers of works in a short period of time have exceeded the sum of all the works released by Sun Yanzi herself since her debut.
In the face of AI Sun Yanzi's prolific cover songs, fans cheered the cover audio to bring them an aural feast, and also began to shout to the singer Buddha, "The rice bowl is risky, hurry up and go out for business."
In response, Sun Yanzi herself responded: "What else do you have to argue with a person who releases a new album every few minutes." ”
AI is developing rapidly, and singers are really at risk of losing their jobs?
In such an era, how should we view the intervention of AI in music production? Good or bad?
AI Music: Already on the way
In fact, the involvement of AI in the production of musical works is not new.
Computer-generated music was already produced in the fifties. In 1951, a programmer produced the first computer-synthesized music in history. In 1957, the string quartet "Ilyak Suite" was composed with the help of an automatic computer.
In the decades that followed, there was an increasing focus on using computers to generate imitations of human musical works, giving rise to virtual singers and enriching AI-empowered music creation.
In 1987, an American composer demonstrated the results of the Musical Intelligence Experiment (EMI), which allowed EMI to extract common creative elements from many man-made compositions and imitate the style of many well-known musicians on this basis.
In 2004, Yamaha Corporation introduced the English version of the speech synthesis engine Vocaloid 1.0, which achieved a breakthrough in synthesizing vocal-style songs.
In 2007, the company's second-generation speech synthesis engine was launched, relying on this technology to launch the virtual idol "Hatsune Miku", whose sound source was sampled by Japanese voice actor Saki Fujita.
Later, Hatsune Miku became one of the hottest virtual idols in the world, and his representative works include "Onion Song", which also led to the subsequent boom in virtual idol development.
Since then, more companies have joined the AI music track. On the one hand, they use artificial intelligence technology to generate a wider variety of music, including game soundtracks, sleep aids, mood music, etc., such as Jukedeck in the United Kingdom created an artificial intelligence system to generate royalty-free music tracks for game developers and enterprises to provide corresponding services.
On the other hand, some companies are also using artificial intelligence technology to provide more convenient tools for music producers, such as the UK's Vochlea company that develops a product to create melodies based on vocal humming.
AI is also more deeply involved in songwriting, such as in 2018, Sony's artificial intelligence composition AI collaborated with human composers to complete the "Hello World" music album, and even its creation began to have emotional elements, such as the AIVA system can create emotional soundtracks for entertainment content.
△ The source of the picture is known
Up to now, we can see that more well-known large companies at home and abroad such as ByteDance, Apple, Google, Amazon, etc. have begun to invest in the AI music field.
It can be seen that the integration of AI technology and music is becoming more and more abundant and diverse, providing more possibilities for future music creation, which is why it is optimistic about capital.
So, how is "AI Sun Yanzi" realized?
The answer given by the UP masters of station B is the "So-vits" algorithm - it can achieve sound cloning. That is, the voice material of a real person is imported, used to train an algorithm, and then through model reasoning, the timbre is converted to generate AI singers.
There are intimate UP masters who have even released AI singer training tutorials.
It can be seen that the biggest feature of AI singers is the timbre of imitation human singers, and "AI Sun Yanzi" can stand out from a group of AI singers, largely due to the uniqueness of the singer's own timbre, but AI singers cannot be perfect in details such as emotions, word bites, and ventilation.
AI impacts the music industry, what do all parties think?
At present, the biggest practical problem brought about by AI's involvement in music production is copyright disputes.
1. The process of generating AI music works involves the risk of infringement
The generation of AI musical works also currently relies on deep learning and training of ready-made human musical works, including copyrighted works, such as those released by pop singers, and the mass copying learning of such works may cross the boundary of "fair use". If this AI is used for commercial profit, it may lead to infringement disputes.
The users of AI tools are also mixed, and the public is often curious and experimented with AI works, but it is easy to ignore the infringement risks therein.
Previously, a B-station UP owner used the portrait of singer Lin Junjie to make multiple AI face-changing videos, and was later sued by the singer for more than 200,000 yuan.
However, a few days ago, Weibo hot search updated the entry of "Zhang Yuanying AI changed face Ye Xiwu". Netizens were delighted, but most did not feel that this move might be an infringement.
2. Whether AI music works have copyright, and its copyright attribution is inconclusive
There are still different views on whether AI musical works have copyright, mainly that AI does not belong to "natural persons and legal persons", and there is also controversy about whether AI musical works are creative.
In addition, the generation of AI music works also involves multiple interest groups, including AI tool developers, users, investors, etc. If AI music works own copyright, there is no clear conclusion on how the copyright income brought by it should be reasonably distributed to each subject.
In the long run, AI singers are out of the circle, and the more difficult it is to distinguish between "real and fake" singers. The involvement of AI in music production also seems to be accelerating the cannibalization of human ingenuity.
So in the face of the impact of AI on music, how have all parties responded?
1. Government: Support new technologies, but improve supervision according to the trend
Countries in various countries are actively promoting relevant regulatory regulations in the field of AI, and have explored the scope of fair use of AI music, the identification of works, and the attribution of rights and interests.
For example, in September 2021, the National New Generation Artificial Intelligence Governance Professional Committee issued the New Generation Artificial Intelligence Ethics Code, which puts forward six basic ethical requirements such as improving human welfare and protecting privacy and security, which are used to guide the healthy development of AI research and development.
The Provisions on the Administration of Deep Synthesis of Internet Information Services, which came into effect on January 10, 2023, defines deep synthesis technology as a technology for generating synthesis algorithms to produce audio and other information, and sets out basic specifications for training data management, content conspicuous identification, and evaluation of technical tools online.
On April 11, 2023, the Cyberspace Administration of China issued a notice for public solicitation of comments on the Measures for the Administration of Generative Artificial Intelligence Services (Draft for Comments), which stipulates that generative AI products include content such as generating sound based on algorithm models, that is, containing current AI music works, and the state supports innovation and application based on AI technology, and at the same time puts forward ethical requirements for products or services extended by such technologies, and the division of responsibilities and obligations between providers and users.
The EU has also issued a proposal for the Artificial Intelligence Law, which proposes the principle of unified AI supervision, and also implements risk grading for AI systems in different application scenarios and proposes regulatory paths respectively.
2. Enterprises: protect copyrights/upgrade service experience/promote industry standards
Different companies have responded differently from their own interests.
(1) Protection of copyright
Music companies have resources such as artists and records.
In the face of AI-generated music, on the one hand, they quickly sounded the alarm, calling for the protection of the copyright of human art creators and their works, and expressed dissatisfaction with AI music infringement, such as for the current online AI-generated songs, Universal Music believes that these songs are not authorized by the original song creators, and tells Spotify and other platforms to prevent AI from grabbing material training from copyrighted songs.
(2) Upgrade services and experience
On the other hand, they are also actively layout, changing with the trend, optimizing services and experiences, and exploring more profit possibilities of AI music.
Foreign streaming platform Anghami tried to launch AI personalized podcasts, while using algorithms to generate music news clips for users.
SoundCloud, an audio social and music communication platform, launched the Fans tool tool based on its own business model to help artists build deeper connections with fans.
At present, domestic QQ Music has launched AIGC vinyl players and other AI-based functional applications to update the user experience.
(3) Promote industry norms
Platform service companies standardize content management based on their own platform characteristics.
The official account of Zhihu issued the "Community Announcement on the Application of AIGC Capabilities to Assist Creation", proposing that the platform will maintain community content by cracking down on the batch release of AIGC content accounts.
In accordance with the Provisions on the Administration of Deep Synthesis of Internet Information Services, Douyin issued the Platform Specification and Industry Initiative on AI-Generated Content, advocating watermarks and metadata specifications for AI-generated content to facilitate users to identify synthetic content.
Compared with the first two platforms that advocate norms and logos, Station B encourages everyone to create creative content and submit articles through AI technology, and also launched the "Virtual Voice Creation Project".
3. Singer: AI = tool / neutral
Faced with the risk that AI may put singers out of work, some people actively embrace change, integrate AI as a tool into their creations, and of course others remain neutral.
Canadian singer Grimes has expressed his willingness to open up his voice rights but asked for a 50% royalty, which directly delineates the distribution of relevant rights.
Taiwanese singer Chen Shanni released a new song "Teach Me How to Be Your Lover", which was tuned by the singer himself and sung by AI. The singer himself expressed his desire to explore the value of original singers in the era of rapid development of AI.
Electronic composer Holly Herndon also steered an artist IP model, where her team created an AI-powered instrument that would allow anyone to sing with her voice, and the profits from the sale of her digital voice were shared by the creator, the DAO and Holly herself.
In the face of the explosion of AI Sun Yanzi, Sun Yanzi herself expressed her unstoppable recognition of the development of AI - "you are not special", "you are already predictable", "you are also customizable", but she still treated it with an indifferent attitude, thinking that it was enough to be herself.
4. People: Entertainment and memes
For AI music works, listeners are more entertaining.
For example, in the comment area of the "AI Sun Yanzi" singing video, the audience seems to be more fond of jokes, thinking that the current AI-generated music works are temporarily inferior to human musicians, and it is better to watch AI singers stimulate real singers to do more business while waiting for technological updates.
AI has come, walk slowly, look forward
The future is here, and there is no way to avoid it. In addition to worry, we may only be able to actively embrace change, and we may wish to look at the good changes brought about by technological updates.
First of all, AI may inspire more excellent works.
As AI's ability to predict and reason songs based on models increases, in addition to assisting human creators to achieve better results, it may also force some "music tailors" on the market to withdraw from the market.
The continuous progress of AI can also urge music creators to think about the differences between humans and AI, give more play to human creativity, use creative expression to inject more soul into their works, convey emotional emotions, and output more in-depth thinking about society and life.
Second, the personalized creation of ordinary people may be realized.
As AI models lower the threshold for use and simplify the operation steps, we may be fortunate to see an era of "music producers for all" - able to meet the needs of personalized creation, so that ordinary people can create songs and express themselves with the help of AI without understanding complex music theory knowledge.
At present, NetEase has made a similar attempt and launched an AI music creation tool - NetEase Tianyin, simplifying the song creation process and providing music creation services for ordinary people.
Finally, when AI intervenes in music production, the deceased art creators who people love may be able to sing again.
For example, at the Jiangsu 2022 New Year's Eve concert, Zhou Shen and Teresa Teng sang on the same stage, and this performance across time and space also used technologies such as AI and AR.
In the future, such similar applications may be more extensive, and it is expected to see more "fan meetings" across time and space.
Finally, I want to say, "But human beings are really immortal."
Walk slowly and look forward.
Reference Articles:
The State of AI Music 2022: Startups, Musicians and Their Challenges
Embodiment, Imagination and Empathy: A Phenomenological Study of Artificial Intelligence Music Generation and Dissemination
""No Man's Music" - From Encoder to Artificial Intelligence Composition of Subject Heterolocation"
"Google AI Music Tool Open Test, UK Launches AI Model Review, Online Performance Market Revenue Nearly 200 Billion Yuan"
"In the era when "AI Sun Yanzi" has become a top stream, how to protect music copyright? 》
"Shenzhen Concludes First Dispute Case of Artificial Intelligence Generated Article Works Dispute, Court Finds Constituent Work"