**The rapid expansion of live virtual people, legal experts analyze the overlooked compliance issues**
The new crown epidemic that lasted for nearly three years has made people accustomed to “online work and study” for a long time, and at the same time has promoted the large-scale and centralized appearance of virtual people to meet people’s “online work and entertainment”.
On June 26, 2021, on the 12th anniversary of the establishment of Station B, Chairman Chen Rui said that in the past year, a total of 32,412 virtual anchors started broadcasting at Station B. Virtual humans play an important role in connecting reality, opening up the metaverse, and becoming an important interactive track in the metaverse.
At the same time, virtual human live broadcasts of merchants have sprung up on major platforms. The “virtual human + story scene” live broadcast has subverted the traditional live broadcast mode, refreshed users’ understanding of platforms and brands, and built a brand new marketing. ecology. In order to achieve sustainable operation, it is necessary for platforms and live broadcast virtual person merchants to pay close attention to the operational and compliance risks involved.
An overview of the development of a live virtual human
The virtual human originated in the 1980s, and in recent years has entered a period of rapid growth with the development of modeling, AI, motion capture, and rendering technologies.
On October 20, 2021, the State Administration of Radio, Film and Television released the “14th Five-Year Plan for Scientific and Technological Development of Radio, Television and Online Audio-visualization”, which proposed in terms of virtual anchors to “explore virtual images such as short video anchors, digital Internet celebrities, and live broadcasts in program interaction. “The application in the link increases the personalization and interest”, which explicitly encourages and supports the development of virtual human for the first time.
“Now that we have entered a point in time, we no longer need to educate customers. I told him why they should push you virtual people. Many brands directly ask which virtual people can cooperate with them. The purpose is very strong, which is also related to the maturity of the market. degree,â? said Ding Yan, founder of Dongtu Universe .
On June 8, 2022, Anycolor was listed on the Tokyo Stock Exchange. This is the world’s first listed company with virtual human as its main business. It has Rainbow Club and live broadcast division.
Now, there have been some deep customization of brand virtual people and companies that provide virtual human IP operation services. Platforms and merchants have begun to customize brand virtual people, or cooperate with companies that provide virtual human IP operation services, trying to use the image of virtual people. The output of IP value will open up its influence in the Z era and expand new consumer groups.
2. Operational risks: live virtual people will also “collapse”
What operational risks will the rapidly advancing live virtual human face?
The virtual human industry chain is divided into basic layer, platform layer and application layer.
– The basic layer includes software and hardware such as modeling software, rendering engines, chips, sensors, optical devices, display devices, etc., and provides a software and hardware basis for the production of virtual humans through 3D modeling, data preprocessing, rendering models , etc. people.
– The platform layer includes the AI â??â??â??â??capability platform, modeling system, motion capture system, rendering system, etc. After obtaining data information from the base layer, the platform layer provides interactive technical capabilities through software algorithms, which can give “life” to virtual humans.
– The application layer creates and operates the virtual human design, shapes the personality, injects “soul” into the virtual human, and applies it to various scenarios.
From the perspective of driving methods, live virtual humans can be divided into non-interactive, human-driven and artificial intelligence-driven.
– Non-interactive refers to responding according to the preset content, and the interaction with the user is extremely weak.
– Human-driven type, also known as “skinned virtual human”, refers to driving virtual human through motion capture equipment based on the actions and expressions of real people.
– Artificial intelligence-driven refers to the use of artificial intelligence (AI) technology to drive virtual human voice expressions, facial expressions, and specific actions in real time or offline through multi-modal technology and the results of deep learning models. â? â? â?
At present, the main applications in the field of live broadcast are human-driven virtual people and artificial intelligence-driven virtual people, both of which can achieve strong interaction with users.
It is generally believed in the industry that there is a risk of the collapse of human settings for live broadcasters, while there is no risk of the collapse of human settings for live broadcast virtual humans. However, the author thinks that the virtual people will also collapse.
“The most important thing for a good virtual person is ‘realism’. It is not only the realism of the image like ‘Liu Yexi’, but also the realism of the personality. He/she can truly cut into the real life and use himself dialogue with consumers,â? said Jojo, Head of Brand Strategy Department of Human Happy Snail Noodles.
Artificial intelligence virtual human is machine learning through a large amount of preset personality-related data, and then automated decision-making through algorithms and interaction with users. The personality modeling of artificial intelligence virtual human may implement the commercial purpose of the platform and merchants, which may lead to the collapse of the personality due to some discrimination and prejudice. The inherent prejudice and discrimination of algorithm designers may also lead to the collapse of human design. For example, a brand of smart speaker recommends suicide when responding to questions about the user’s heart cycle, and an artificial intelligence software identifies blacks as primates.
In February 2022, Article 21 of the “Internet Diagnosis and Treatment Supervision Regulations (Trial)” jointly issued by the National Health Commission and the State Administration of Traditional Chinese Medicine emphasized that “it is strictly forbidden to use artificial intelligence to automatically generate prescriptions, and it is strictly forbidden to use artificial intelligence for commercial purposes. The reason is that it is difficult to use artificial intelligence to automatically generate prescriptions without reflecting the commercial purposes of the platform and merchants.
The skinned virtual human first needs to create an avatar, commonly known as a “holster”; then the “soul” injection of the skinned virtual human is realized by a real person (commonly known as “the person in the middle”) who provides the virtual human with voice and action, because “The Man in the Middle” is not only responsible for the voice performance, but also substituting his own personal emotions and preferences into it, shouldering the important task of giving the character “soul”.
Many people think that as long as they don’t “collapse” because of negative news like real idols, the life cycle of virtual idols is infinite. In fact, once the virtual idol is replaced by the “people in the middle”, the original emotional connection with the fans will come to an abrupt end, and the life cycle of the character will come to an end.
But whether it’s an AI live broadcast or a character set completely designed according to the script, it’s not what fans really want. Capital may think that fans love leather cases, but for many fans, the emotional value provided by virtual idols is more important than cold leather cases. The holster and the soul complement each other. Fans must be attracted by the holster at first sight, but an interesting soul is the key to maintaining popularity.
Compliance Issues Faced by Three Human-Driven Virtual Humans
Although virtual idols are technological products, they still have to be driven by people.
“Fans empathize their love for virtual idols to the “people in the middle” (the people behind the virtual idols) behind them, creating chaos between virtual and reality. The hibernation incident of A-SOUL members is actually caused by this chaos Chaos, in which the content of justice is worth pondering.â?
The above comments from People’s Daily Online exemplify the current situation of virtual human beings. The holster is the character design of the perfect script of the virtual human set by the platform or the business. However, even if the “Man in the middle” behind the holster is completely performed according to the character design of the perfect script, different “Man in the middle” will still Unconsciously interpreting different styles on the holster is to reflect the different souls of virtual people, just like the same movie script, different actors have different interpretations of the same role, and ultimately reflect different styles.
So, what are the respective compliance risks and responsibilities of the platform and the merchant for the avatars?
(1) Platform
For a platform that gathers a large number of live broadcast virtual people, the platform’s good management obligations are positively related to the economic benefits directly obtained from the virtual people’s live broadcast (for example, sharing in the rewards obtained by the live broadcast virtual people) , the more economic benefits obtained. Large, naturally should have a higher duty of care for this.
1. Conduct a preliminary review of the copyright of the live avatar image (holster). Platforms not only need to fulfill the responsibilities of general network service providers, but also should establish a set of intellectual property review mechanisms to conduct preliminary review of the copyright aspects of live avatar images on the platform. This review should adopt the “general possibility” standard, for example, review the preliminary evidence such as copyright manuscripts, originals, legal publications, copyright registration certificates, certificates issued by certification agencies, etc., to prove that they are the owners of copyrights or copyright-related rights and interests. In short, platforms cannot abuse the “notify-take-down” rule.
2. The platform must fulfill the necessary audit obligations. When opening the online tipping function, the platform should require merchants to provide relevant identity certificates, qualification certificates for engaging in related industries, products or services and other documents, and fulfill their necessary audit obligations . After receiving relevant complaints, the platform should verify and verify in a timely manner. If there is indeed a problem with the live broadcast virtual person, it should take a series of measures such as freezing the account to avoid further expansion of user losses.
3. The platform should highly guard against possible fraudulent behaviors by merchants. For example, the Vox Akuma live-streaming virtual human claims, “Vox was a Japanese demon who lived 400 years ago, and was later destroyed by the Tokugawa family. Because of the blood of the demon, Vox was resurrected 400 years later. I hope to understand modern times through the Internet. society and rebuilding the ethnic group”.The “Resurrection and Reconstruction of the Family” created by Vox Akuma won a reward of 1.11 million yuan in the live broadcast of more than an hour. This reward is different from the general reward for pure appreciation. It may also be that everyone really wants to help Vox rebuild. For ethnic groups, the use of this fund needs to be cautious to avoid suspected fraud and fraud, especially in the case that the platform takes part of the reward income from it, the platform may need to take greater responsibility.
4. The platform needs to pay close attention to joint and several liabilities for infringement of civil rights and interests. Article 1197 of the “Civil Code” stipulates the joint and several liabilities of network service providers. When the platform takes part of the income from the reward, it is likely to constitute “the network service provider knows or should know about the network Users use their network services to infringe on the civil rights and interests of others”, in the case of not taking necessary measures, the platform will be jointly and severally liable with the merchants.
(2) Merchant
1. Intellectual property infringement. To customize a live broadcast virtual person, a merchant should first obtain the copyright of the live broadcast virtual person image (holster) and the complete human design concept. The copyright chain should be clear, and at least it should be authorized to include distribution rights, information network dissemination rights, modification rights, reproduction rights, performance rights, and adaptation rights, and must not infringe on the prior rights of others.Secondly, before obtaining data such as voices, movements, facial expressions, etc., the consent of the “Zhongzhiren” should be obtained, and a relevant confidentiality agreement should be signed with the “Zhongzhiren”, and the relevant information of the “Zhongzhiren” ” should not be disclosed to meet the needs of the “Zhongzhiren”. The completeness and fullness of the live avatars.
2. The problem of algorithmic discrimination. Merchants cooperate with companies that provide virtual human IP operation services. They must clarify the rights and obligations of both parties. They must not implement their business values â??â??â??â??in the algorithm, and must not bring discrimination and prejudice into the algorithm. Maintain the integrity and integrity of the live virtual human design. full.
3. The setting of the live virtual human must conform to scientific ethics, and ensure that the technology is improved for the good.
Four Compliance Issues Faced by Artificial Intelligence Virtual Humans
Intelligent virtual human refers to the use of artificial intelligence technology, through multi-modal technology and the calculation results of deep learning models, real-time or offline driving virtual human voice expression, facial expressions, specific actions, it uses artificial intelligence technology to create a complete live broadcast virtual human. Character design, and endow the live virtual person with a “soul”.
The use of artificial intelligence should comply with the principles of comprehension and fairness, and artificial intelligence should be able to be understood by the existing morality of human society; the operation of artificial intelligence should maintain the principle of transparency and visibility, and prevent specific individuals or minority groups from being prejudiced, humiliated, and humiliated. Discrimination; AI should respect human dignity and personal privacy, and protect AI participants’ right to know and choose; finally, the development of AI should be controllable, and human beings should not be subordinated to AI, nor should they be coerced by AI. To ensure the decisive role of human beings in artificial intelligence, when artificial intelligence is out of control, there should be a destruction device.
For the live broadcast operation of intelligent virtual humans, the compliance risks that platforms and merchants need to pay attention to are as follows:
(1) Platform
For a platform that gathers a large number of live broadcast virtual people, the platform’s good management obligation is positively related to the economic benefits directly obtained from the live broadcast of virtual people (for example, sharing in the rewards obtained by the live broadcast virtual person ), the economic benefits obtained The greater the benefit, the higher the duty of care should naturally be, and the legal responsibility and duty of care that it should undertake is basically the same as the human-driven virtual human described earlier.
(2) Merchant
1. When producing and operating live virtual humans, scientific ethics should be adhered to first.
Ethics refers to the principles and principles that should be followed when dealing with the interrelationships between people and between people and society. It not only contains behavioral norms when dealing with the relationship between people, people and society, and people and nature, but also contains behavioral norms profound truths to regulate behavior according to certain principles.
The biggest weakness of artificial intelligence is that it lacks the ability to directly feel and has weaknesses in value judgment. The value concepts of benefiting human beings, avoiding harm, fairness and justice advocated by the ethical standards of artificial intelligence provide the development and application of artificial intelligence technology. value judgment criteria.
The ethical outlook on the development and application of artificial intelligence should meet the following goals:
– We should protect human dignity, integrity, freedom, privacy and safety in society and at work for the common interests and human well-being, with the purpose of promoting human beings to improve their lives, promote the fair sharing of the benefits of artificial intelligence in the whole society, and continuously strengthen the people’s the sense of gain, happiness and security, and promote the peaceful development and sustainable development of human society;
– The operation should conform to the principles of comprehensibility and fairness. Artificial intelligence should be able to be understood by the existing morality of human society, effectively protect the legitimate rights and interests of all relevant subjects, and promote social fairness, justice and equal opportunities;
– The operation of artificial intelligence should maintain the principle of transparency and visibility, and not subject specific individuals or minority groups to prejudice, stigma and discrimination;
– Respect the right to life, artificial intelligence should avoid harm or potential threats to human life safety, physical health, mental and mental health to the greatest extent, respect personal dignity and personal privacy, and protect the right to know and the right of choice for AI participants;
– The development of artificial intelligence should be controllable. Humans should not be subordinated to artificial intelligence, nor should they be coerced by artificial intelligence, so as to ensure the decisive role of human beings in artificial intelligence. When artificial intelligence is out of control, there should be a destruction device.
2. Data ethics
Artificial intelligence technology represented by machine learning requires a large amount of data. Data is the foundation of machine learning. The authenticity and integrity of data, as well as the scale and quality of data are the keys to artificial intelligence analysis and modeling.
The core foundation of AI fairness is data. The unfairness of data is mainly reflected in the following: the unbalanced proportion of different groups in the data, the inherent discrimination in the data, the inherent bias in the data, the inclusivity of the data is not enough, and the generality of the data is not enough. data accuracy, etc.
Data is inherently discriminatory and biased because some discriminatory and biased values â??â??â??â??are mapped onto the data. If the above-mentioned data unfairness is not debugged and verified, this unfairness will be fed back to the AI â??â??â??â??algorithm model and finally mapped to the fairness of AI.
3. Algorithmic governance
Algorithm is a “data processing activity”, including collecting and processing data, mining data value, covering the whole process of data generation, collection, storage, transformation, processing and output.
The intelligent live virtual human needs to realize automatic decision-making through algorithms to achieve strong interaction with users.
The emergence and use of algorithms is to perform large-scale repetitive tasks and calculations, to achieve efficient and fast commercial transactions, and to achieve the breadth, depth and efficiency of the operation of public power.
With the increasingly widespread application of algorithms, algorithms have become a tool to promote political, economic, cultural and social development. They should follow the common values â??â??â??â??of mankind and abide by the ethics and morality of the country or region. Algorithms should respect the dominant position of human beings and give users the right to choose and make decisions. ; Algorithms should suppress ubiquitous discrimination and prejudice in human society; Algorithms should protect users’ personal and property rights.
https://www.36kr.com/p/1783406419624322