On September 25, 2021, the National New Generation Artificial Intelligence Governance Professional Committee issued the "New Generation Artificial Intelligence Ethical Specifications" (hereinafter referred to as the "Specification"), which made specific requirements on user exit mechanisms, algorithmic discrimination and false propaganda, It responds to the current ethical concerns of all sectors of society regarding privacy, prejudice, discrimination, and fairness, and aims to integrate ethics into the entire life cycle of artificial intelligence.

Experts told Nandu AI Outpost that the issues mentioned in the "Specification" have been reflected in the Personal Information Protection Law and the "Regulations on Several Issues Concerning the Application of Laws in the Trial of Civil Cases Related to Use Face Recognition Technology to Handle Personal Information" , One of the key points is to protect users' right to choose.

One of the key points is to protect users' right to choose

The "Normative" proposes that all kinds of artificial intelligence activities should follow six basic ethical norms: improving human welfare, promoting fairness and justice, protecting privacy and security, ensuring controllability and credibility, strengthening responsibility, and improving ethical literacy.

Article 4 of the "Normative" mentions that humans should be guaranteed to have full independent decision-making power, have the right to choose whether to accept services provided by artificial intelligence, have the right to withdraw from interaction with artificial intelligence at any time, and have the right to suspend the operation of artificial intelligence systems at any time to ensure that Artificial intelligence is always under human control.

Previously, the "Regulations on the Recommendation Management of Internet Information Service Algorithm (Draft for Comments)" issued by the State Cyberspace Administration of China also proposed relevant measures for personalized recommendations, and proposed to require algorithm recommendation service providers not to record illegal and bad information keywords into users Discriminatory or prejudicial user tags shall not be set up when points of interest are used as user tags and pushed information accordingly.

In this regard, Duan Weiwen, director of the Center for Science and Technology and Social Research of the Chinese Academy of Social Sciences, believes that one of the key points in the issue of artificial intelligence services is to protect users' right to choose. This means that in the future, the field of artificial intelligence application must respect the user's right to choose. If users refuse to accept artificial intelligence services, they need to provide alternative traditional technical services.

He gave an example: "Taking human customer service as an example, companies can decide which projects to provide human customer service other than artificial intelligence based on actual efficiency, especially when financial transactions are involved, companies need to make corresponding supplements."

In addition, Duan Weiwen believes that artificial intelligence should respond to the actual standards in terms of technical measures, such as setting up pop-up reminders on the interactive interface to remind users that they can choose to exit the artificial intelligence service. "In the development of new technologies, the public's acceptance of them is processed, and this process cannot affect the actual interests of users. The dividends of artificial intelligence should be considered as much as possible for most users," he said.

Let high-quality data flow

Nandu AI Outpost noticed that in Article 11, the "Specification" has made data integrity, timeliness, consistency and specifications on the collection, storage, use, processing, transmission, provision, and disclosure of data. requirements such as sex and accuracy.

Duan Weiwen believes that this part of the regulations is related to many specific practical issues. The data included include personal information data, and also includes data in manufacturing applications. The problem of data quality is closely related to digital transformation.

"This involves how to make high-quality data flow. As of now, the reason why some artificial intelligence applications are not of high quality is actually related to the quality of the data." Duan Weiwen said, "If it is in the process of data collection, if it is in the process of data collection," said Duan Weiwen. Not representative of the data may lead to discrimination and prejudice. From the perspective of ethical norms, data with problematic quality may have negative social impacts or ethical risks."

How to improve data quality? Duan Weiwen believes that on the one hand, the artificial intelligence industry should fully recognize the importance of data in the artificial intelligence industry and strengthen the universality and representativeness of sample data; on the other hand, it is necessary to strengthen data sharing and interoperability to overcome the problem of data silos.

False propaganda issues will be managed consistently

Regarding the limitations of artificial intelligence services, Article 16 of the "Specifications" proposes: "Using artificial intelligence technology in products and services should be clearly informed, and the functions and limitations of artificial intelligence products and services should be identified to ensure that users are informed and agreed to ”. This limits the current phenomenon of false propaganda and exaggeration of product functions in artificial intelligence products.

Previously, Tesla was investigated for suspected illegal publicity in the United States. Some time ago, NIO's publicity also caused public controversy. Does the focus of car companies’ expressions in publicity and marketing involve false marketing? How should this phenomenon be regulated?

Duan Weiwen believes that in artificial intelligence applications, users’ right to consult and consent is required. In the application of artificial intelligence, some problems are gradually exposed during use.

"If we look at it from a development perspective, artificial intelligence was the object of hype in the previous stage. But with the promotion and use, and the gradual maturity of users, the limitations of the product will gradually emerge. This does involve false propaganda. "Duan Weiwen said, "After the release of the "Specification", there should be consistent management of this issue."

However, such management needs to be refined into ethical standards for services in various industries in order to implement relevant ethical norms. "The current norms mainly serve as reminders. By proposing specific product and service ethics standards for the corresponding industries, this issue can be improved," said Duan Weiwen.

He also mentioned that in addition to the determination of regulatory authorities to implement it, the power of social supervision is also needed. "For example, China's social supervision requires the public to pay attention to these issues, and in addition, media think tanks should also play a role in promoting it."

Interview and writing: Nandu intern reporter Hu Gengshuo