Overview of 6G AI Data Services

With the continuous maturity of AI technology and the continuous enrichment of application scenarios, AI applications represented by deep learning have radiated from pan C-end fields such as consumption and the Internet to traditional industries such as manufacturing, security and medical care. The large-scale implementation of AI technology innovation and application has driven the vigorous development of the big data intelligent market, and also injected market vitality into data services. 5G has opened up the scene connection of the Internet of Things, connecting thousands of industries to mobile communication networks, bringing new scenarios and ubiquitous data. One of the functions of 6G networks is to empower AI capabilities to applications and scenarios in various fields based on ubiquitous big data, creating an "intelligent ubiquitous" world through wide area coverage and intelligent adaptation to scenarios. Therefore, 6G networks require the construction of endogenous, ubiquitous, and distributed AI capabilities [1]. In the 6G network, endogenous intelligence can utilize AI algorithms internally to improve network performance, enhance user experience and efficiency, and provide various support capabilities for AI externally, making AI training/inference more efficient and real-time. It can extract and encapsulate network intelligence, serving customers in different industries. There will be a wide variety of AI model data and training data in the new 6G system, which need to be able to interact and share among various network element nodes in different network architectures. With the commercialization of the AI industry, the demand for more forward-looking dataset products and highly customized data services will become mainstream.

Mobile communication networks, as producers and consumers of data, are transforming from carriers of data to data platforms with trustworthy control and value mining. The existing communication networks represented by 5G serve as the "pipeline" for communication session data transmission, only playing the role of establishing information exchange paths between terminal devices and networks. But as endogenous perception and intelligence become the fundamental capabilities of 6G networks, distributed and ubiquitous computing power and data become new network service capabilities beyond communication connections. Unlike point-to-point transmission of communication session data, the challenge faced by processing intelligent, perceptual, and network operation generated and consumed data is the need for a distributed method for data collection, preprocessing, storage, and analysis. To this end, it is necessary to design a data service architecture independent of traditional user side, systematically solving the challenges of 6G mobile communication networks in controlling and realizing the value of non user side data.

The Value and Potential of 6G AI Data Services
Endogenous perception and intelligence will be the two main new capabilities of 6G networks. The former generates massive amounts of data, while the latter makes automatic decisions based on data to improve network performance or provide services to upper level applications. Thanks to the tremendous success of big data technology and artificial intelligence, its application in wireless networks is becoming increasingly mature. According to Huawei's forecast, the total annual data generated globally by 2030
The quantity will reach 1YB, an increase of 23 times compared to 2020. The realization of the value of a large amount of data requires the use of intelligent analysis algorithms, and the true value of data depends on its liquidity and data collaboration. Therefore, privacy protection and open sharing of data are important mechanisms for achieving data flow and reflecting value; The owner or provider of data can only realize value realization by providing data to data consumers in the form of a service. Unlike the general realization of asset value, data realization is not a one-time transaction and requires
To ensure the realization of duplicate value realization of data under privacy protection. For network AI, an AI model trained on a large amount of training data is the value presentation of a large amount of data after algorithm mining, and it is also an important intellectual property that needs to be protected.