Author | Wang Zhasha Editor | Fan Zhihui
"2023 is the year when artificial intelligence technology moves out of the field of science fiction."
Faced with the rapid progress of artificial intelligence technology last year, billboard commented this way. The development of
technology is also forcing the improvement of the system. At the end of the year and the beginning of the year, the field of artificial intelligence ushered in a wave of "legislation". In the past month, Europe and the United States have successively proposed artificial intelligence regulatory bills, many of which are closely related to the music industry.
On December 8 last year, the European Parliament, EU member states and the European Commission reached an agreement on the Artificial Intelligence Act. On January 10, a bipartisan group of House representatives led by Rep. Madeleine Dean, a Pennsylvania Democrat, and Maria Salazar, a Florida Republican, proposed the Artificial Intelligence Fraud Prohibition Act, which aims to protect people’s images. and voices from AI-generated deepfakes.
Through these "French words", we can also see the legislators' stance and response to some controversial events that occurred in the music industry in the past year.
“Artificial intelligence platforms should not receive a free pass”
After the EU reached an agreement on the Artificial Intelligence Act, European Commission President von der Leyen posted on social media: The Act is the first of its kind in the world and sets up a unique framework for the development of artificial intelligence. legal framework.
Although the EU has not yet announced the final version of the "Act", according to the disclosed proposal information, the "Act" has formulated a regulatory framework around the development, market launch and use of artificial intelligence systems, adopting a risk-based regulatory approach and providing different Risk levels impose different requirements and obligations on AI systems.
For example, for artificial intelligence that is “intrusive” or “discriminatory”, or that exploits the weaknesses of individuals or specific groups, substantially distorts human behavior, and causes or may cause significant harm to human beings, the Act It will be regarded as "artificial intelligence systems that pose unacceptable risks to human security" and the development and use of such systems will be prohibited. For systems that are allowed to be developed and used, the "Act" provides principle provisions such as "minimizing accidental harm and maintaining robustness when problems arise" and "respecting human dignity and personal autonomy" to ensure that AI technology has sequential development.
The music industry is an important "test field" for AI technology. All parties in the music industry have expressed their demands for the disputes and problems that arise during the practice of AI.
Before the EU reached an agreement on the Artificial Intelligence Bill, many organizations in the music industry had launched several calls, hoping that the EU would strengthen the protection of the rights of musicians in the bill. It is worth noting that in addition to regular expressions such as calling for strengthening copyright protection, the word "transparency" appeared in the speeches of several industry giants.
Bjorn Ulvaeus, founder of ABA and president of the International Confederation of Societies of Authors and Composers (CISAC), reminded policymakers in a letter to EU ambassadors that without transparency By making a request, the creator's rights will be compromised.
requires artificial intelligence service providers to fulfill transparency obligations. In layman's terms, it means that they need to explain what content is used in the AI pre-training process. If the AI system has enough transparency, it will be difficult for providers to use unauthorized sounds, lyrics and music content to train artificial intelligence without anyone knowing.
can see that the Artificial Intelligence Act does indeed reflect this requirement. In the current version of the bill, “artificial intelligence (gpai) systems and the gpai models on which they are based must comply with transparency requirements,” according to the European Parliament website.
Compared with the European Union's Artificial Intelligence Act, the proposal of the United States' Artificial Intelligence Fraud Prohibition Act has more specific direct reasons and its direction is more obvious. According to reports, representatives sponsoring the bill released a fact sheet that stated: "Unscrupulous companies and individuals are 'hijacking' the voices and likenesses of professionals, undermining the legitimate works and aspirations of important contributors to American culture and commerce."The
manual also specifically mentioned the song "heart on my sleeve" released by ghostwriter977 in April last year. This "creator" used the voices of well-known musicians drake and the weeknd to edit the AI model without authorization. After training, the song was finally generated and attracted widespread attention on social media.
At the same time, Universal Music, the record company owned by Drake and The Weeknd, immediately complained, and platforms such as Spotify, Apple Music, YouTube, and TikTok quickly removed the song. After this incident, the music industry saw a potential crisis. The National Music Publishers Association sent a letter to the U.S. Senate majority party stating that "artificial intelligence platforms should not receive a free pass."
After half a year, the "Artificial Intelligence Fraud Prohibition Act" mentioned the case of "heart on my sleeve" again. Obviously, the new bill is highly targeted for this type of AI songs generated using unauthorized materials. The principle has been established at the national level that "everyone has intellectual property rights over his or her own image and voice, allowing individuals to seek monetary compensation for harmful or unauthorized use of their image or voice." Things like AI-generated fake drake will also be clarified prohibit.
In addition, the state of Tennessee in the United States also announced a state bill called the "Ensuring Similarity Voice and Image Security Act" on the 10th of this month. The legislative intent of this bill is similar to the federal "Artificial Intelligence Fraud Prohibition Act." The bill emphasizes Using artificial intelligence to generate false recordings without authorization will be subject to legal action in Tennessee.
The music industry’s anxiety about the prospects of AI technology and its calls for relevant laws are understandable.
As the Billboard review article said, artificial intelligence technology is "rudely breaking into daily life". Prompt and reasonable regulation at the institutional level can minimize the negative impact of artificial intelligence technology and prevent the music industry from becoming A primitive jungle of chaos and disorder, where the weak and the strong prey on each other.
Why is AI regulation and legislation important?
According to the "Music Participation Report 2023" survey released by ifpi, 76% of respondents believe that artificial intelligence should not use or ingest artists' music or singing without permission, and 74% of respondents believe that artificial intelligence should not use or ingest artists' music or singing without permission. Should be used for cloning or unauthorized impersonation of artists.
But it is undeniable that before the announcement of the Artificial Intelligence Act, various industries did not reach a unified consensus on copyright issues related to artificial intelligence.
Last spring, "ai Stefanie Sun" became popular all over the Internet. Based on an open source project called so-vits-svc, a generative model can be used to synthesize the audio of the target timbre using only a few pieces of audio, and train the acoustic model that the user wants. Compared with the popularity of second-generation creative activities, few people are interested in topics such as "Whether AI Stefanie's use of Stefanie Sun's works and voices for training is an infringement?"
There are also large differences around the world regarding the criteria for determining artificial intelligence infringement. While many people and organizations have repeatedly publicly called for artificial intelligence systems to respect intellectual property rights, there are also many people who play the role of "Cyber Kong Yiji" and think "can artificial intelligence be considered stealing?"
last year In November, Ed Newton-Rex, head of audio at technology company Stability AI, resigned. The industry leader, who has worked in the fields of artificial intelligence and music for 13 years, said that he left because of disagreement with the company's philosophy. Newton-rex believes that any artificial intelligence developer's use of the creative work of others without consent is "exploitative," but Stability AI feels that acquiring copyrighted content is "fair use." In an interview with the media, Newton-Rex also specifically emphasized that "this is true for most companies."
Even some areas at the government level tend to exempt the unauthorized use of materials to train AI. In May 2023, Japanese Minister of Education, Culture, Sports, Science and Technology Keiko Nagaoka stated that Japanese law will not protect copyrighted materials used in the AIGC model pre-training set.
However, with the EU's positive progress in artificial intelligence legislation, relevant companies can no longer use the wisdom of others without authorization in the name of "technological innovation" and "resource sharing".At the level of law enforcement that is more specific, clear and operable, European and American bills have also updated and improved the existing legal framework, making the determination of infringements simpler and clearer, and reducing the cost of rights protection.
takes the U.S. Artificial Intelligence Fraud Prohibition Act as an example. Prior to the act, only 19 of the 50 U.S. states had laws explicitly recognizing some form of publicity right. It is understood that the right of publicity is an intellectual property right that protects against unauthorized use of a person's likeness, voice, or other aspects of their identity. Additionally, 11 states recognize the right of publicity only as a common law.
In other words, before the promulgation of the Artificial Intelligence Fraud Prohibition Act, courts in some states in the United States may consider "sound" not to be a copyright.
"Bloomberg Law" pointed out that although the song "Heart on My Sleeve" recorded by the AI-generated fake Drake and The Weeknd was deleted from the streaming platform due to Universal Music's rights protection, this was probably because of this song. Contains audio samples of other songs, which contain the well-known tag of producer metro boomin, and the existence of this tag constitutes infringement.
In other words, if there is no such label, and the voices of drake and the weeknd are unauthorized, under the legal framework at that time, this rights protection claim may not be supported by the court.
In the context of the rapid development of AI technology, the legal lag in the "right to voice" is actually a problem that exists widely around the world.
In June last year, the Japan Actors Guild also announced the "Proposal on the Use of Generative Artificial Intelligence Technology", calling on the industry and the government to review the operation of copyright laws and formulate rules to establish "voice portrait rights." According to reports, in Japan, AI-generated images of actresses and female voice actors have appeared, and there are even unofficially authorized AI voice actor voice changers for sale. Therefore, relevant legislation needs to be improved urgently.
In our country, according to Article 1023 of the Civil Code, the protection of natural persons’ voices shall be governed by the relevant provisions on the protection of portrait rights. In the country's first dispute over the rights to the lines in a film and television drama, a software developer and supplier of a certain game was ordered to sue the plaintiff for the unauthorized use of lines from a classic segment of the TV series "Conquest" starring actor Sun Honglei. Apologize and compensate for economic losses of 30,000 yuan.
It is worth mentioning that domestic legislation on artificial intelligence has been promoted in an efficient and orderly manner. On April 11 last year, the Cyberspace Administration of China took the lead in issuing a public opinion solicitation notice on the "Measures for the Administration of Generative Artificial Intelligence Services (Draft for Comment)". The "State Council 2023 Legislative Work Plan" announced in June included artificial intelligence The draft law is included in the legislative items to be submitted to the Standing Committee of the National People's Congress for review, which means that national-level artificial intelligence legislation has entered a new stage.
However, some technology companies that have criticized the AI control bill said that although the formulation of various new rules protects the rights and interests of copyright holders, it also makes technology developers more constrained and slows down the pace of innovation. In the long run, It is detrimental to human development.
But in fact, reasonable and effective supervision will do more good than harm to artificial intelligence technology. There is a saying in jurisprudence that "determination of disputes ends", which means that only by reducing the uncertainty about the ownership of rights can further transactions and distributions be carried out.
In the face of the development of artificial intelligence technology, many musicians hope to participate. Facts have also proved that the active participation of musicians can help the artificial intelligence team calibrate and optimize the project.
Last year, artificial intelligence voice company hooky contacted singer lauv, and the two parties collaborated to translate lauv's new single "love u like that" into a Korean version. lauv was responsible for providing his original vocals in different ranges. For the AI team, the sound materials provided by musicians who proactively cooperate with the project will definitely meet the needs of the project better than the materials they grab from the Internet without authorization.
When asked if he would be willing to provide the voice model for public use, lauv's attitude was very positive, saying: "As long as the whole world knows which songs I wrote and produced... I don't have any excuse. Don't do this."
When the relevant bills clarify the ownership of rights, the enthusiasm of singers, record companies and other rights holders will increase in the process of cooperating with AI music projects, and they will no longer worry about being infringed and having no way to complain. Artificial intelligence service providers have obtained After authorization, the copyright risk in the later stages of the project will be greatly reduced, and you can be free to carry out technology research and development.
is from a technical perspective. For musicians, is artificial intelligence more like a convenient tool or a tool that may replace themselves? This may not be easy to say for competitors. But at the institutional level, clarifying the simple legal principle that "if you want to use other people's things, you must obtain the consent of others in advance" will definitely have more advantages than disadvantages, and it will help to form a relationship between copyright owners and artificial intelligence. A win-win situation for smart technology holders.
Conclusion
At the recent Davos Forum, United Nations Secretary-General Guterres pointed out: “Every new breakthrough in generative artificial intelligence adds unexpected threats. . "
But technological trends are irreversible, and we cannot stop eating because of choking. As Zhao Tingyang, a member of the Chinese Academy of Social Sciences, once proposed, "Human beings can pre-design a kind heart for super artificial intelligence." This is undoubtedly inseparable from the legislative level reached by countries around the world. Interoperability and consensus, including the efforts made by China, the United States, the European Union, etc. in regulating artificial intelligence, are to implant a kind heart into artificial intelligence.
Although this kind heart implantation operation has not been completely successful, it will always be found The best balance between risk control and technological development.
Although the road is far away, the journey is coming.
Typesetting | a
